diff --git "a/attnserver.run_attnserver.slurm.sh.343241.err.log" "b/attnserver.run_attnserver.slurm.sh.343241.err.log" new file mode 100644--- /dev/null +++ "b/attnserver.run_attnserver.slurm.sh.343241.err.log" @@ -0,0 +1,5105 @@ ++ source /mnt/weka/home/hao.zhang/conda/miniconda/bin/activate +++ _CONDA_ROOT=/mnt/weka/home/hao.zhang/conda/miniconda +++ . /mnt/weka/home/hao.zhang/conda/miniconda/etc/profile.d/conda.sh ++++ export CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda ++++ CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda ++++ export _CE_M= ++++ _CE_M= ++++ export _CE_CONDA= ++++ _CE_CONDA= ++++ export CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python ++++ CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python ++++ '[' -z x ']' +++ conda activate +++ local cmd=activate +++ case "$cmd" in +++ __conda_activate activate +++ '[' -n '' ']' +++ local ask_conda ++++ PS1= ++++ __conda_exe shell.posix activate ++++ '[' -n '' ']' ++++ /mnt/weka/home/hao.zhang/conda/miniconda/bin/conda shell.posix activate +++ ask_conda='unset _CE_M +unset _CE_CONDA +PS1='\''(base) '\'' +export PATH='\''/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin'\'' +export CONDA_SHLVL='\''1'\'' +export CONDA_PROMPT_MODIFIER='\''(base) '\'' +export CONDA_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda'\'' +export CONDA_PYTHON_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/python'\''' +++ eval 'unset _CE_M +unset _CE_CONDA +PS1='\''(base) '\'' +export PATH='\''/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin'\'' +export CONDA_SHLVL='\''1'\'' +export CONDA_PROMPT_MODIFIER='\''(base) '\'' +export CONDA_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda'\'' +export CONDA_PYTHON_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/python'\''' ++++ unset _CE_M ++++ unset _CE_CONDA ++++ PS1='(base) ' ++++ export PATH=/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin ++++ PATH=/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin ++++ export CONDA_SHLVL=1 ++++ CONDA_SHLVL=1 ++++ export 'CONDA_PROMPT_MODIFIER=(base) ' ++++ CONDA_PROMPT_MODIFIER='(base) ' ++++ export CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda ++++ CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda ++++ export CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python ++++ CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python +++ __conda_hashr +++ '[' -n '' ']' +++ '[' -n '' ']' +++ hash -r ++ conda activate junda-attnserver ++ local cmd=activate ++ case "$cmd" in ++ __conda_activate activate junda-attnserver ++ '[' -n '' ']' ++ local ask_conda +++ PS1='(base) ' +++ __conda_exe shell.posix activate junda-attnserver +++ '[' -n '' ']' +++ /mnt/weka/home/hao.zhang/conda/miniconda/bin/conda shell.posix activate junda-attnserver ++ ask_conda='unset _CE_M +unset _CE_CONDA +PS1='\''(junda-attnserver) '\'' +export PATH='\''/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin'\'' +export CONDA_PREFIX='\''/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver'\'' +export CONDA_SHLVL='\''2'\'' +export CONDA_DEFAULT_ENV='\''junda-attnserver'\'' +export CONDA_PROMPT_MODIFIER='\''(junda-attnserver) '\'' +export CONDA_PREFIX_1='\''/mnt/weka/home/hao.zhang/conda/miniconda'\'' +export CONDA_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda'\'' +export CONDA_PYTHON_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/python'\''' ++ eval 'unset _CE_M +unset _CE_CONDA +PS1='\''(junda-attnserver) '\'' +export PATH='\''/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin'\'' +export CONDA_PREFIX='\''/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver'\'' +export CONDA_SHLVL='\''2'\'' +export CONDA_DEFAULT_ENV='\''junda-attnserver'\'' +export CONDA_PROMPT_MODIFIER='\''(junda-attnserver) '\'' +export CONDA_PREFIX_1='\''/mnt/weka/home/hao.zhang/conda/miniconda'\'' +export CONDA_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda'\'' +export CONDA_PYTHON_EXE='\''/mnt/weka/home/hao.zhang/conda/miniconda/bin/python'\''' +++ unset _CE_M +++ unset _CE_CONDA +++ PS1='(junda-attnserver) ' +++ export PATH=/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin +++ PATH=/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/.local/bin:/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin:/mnt/weka/home/hao.zhang/conda/miniconda/condabin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin +++ export CONDA_PREFIX=/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver +++ CONDA_PREFIX=/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver +++ export CONDA_SHLVL=2 +++ CONDA_SHLVL=2 +++ export CONDA_DEFAULT_ENV=junda-attnserver +++ CONDA_DEFAULT_ENV=junda-attnserver +++ export 'CONDA_PROMPT_MODIFIER=(junda-attnserver) ' +++ CONDA_PROMPT_MODIFIER='(junda-attnserver) ' +++ export CONDA_PREFIX_1=/mnt/weka/home/hao.zhang/conda/miniconda +++ CONDA_PREFIX_1=/mnt/weka/home/hao.zhang/conda/miniconda +++ export CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda +++ CONDA_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/conda +++ export CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python +++ CONDA_PYTHON_EXE=/mnt/weka/home/hao.zhang/conda/miniconda/bin/python ++ __conda_hashr ++ '[' -n '' ']' ++ '[' -n '' ']' ++ hash -r ++ export CHROME_TRACE_PREFIX=/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5 ++ CHROME_TRACE_PREFIX=/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5 ++ mkdir -p /mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5 ++ export PROF_TP_SIZE=2 ++ PROF_TP_SIZE=2 ++ export PROF_CP_SIZE=8 ++ PROF_CP_SIZE=8 ++ export PROF_BS=16 ++ PROF_BS=16 ++ for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072 ++ export PROF_CTX_LENGTH=1024 ++ PROF_CTX_LENGTH=1024 ++ name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L1024*tp2.cp8.bs16.json' ++ '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L1024*tp2.cp8.bs16.json' ']' ++ echo 'Running ctx_length=1024, TP_SIZE=2, CP_SIZE=8, BATCH_SIZE=16' ++ srun bash ./attnserver.sh ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 0 --rdzv_id 343241 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-239:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 1024 --max-position-embeddings 1024 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 1 --rdzv_id 343241 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-239:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 1024 --max-position-embeddings 1024 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:11:27.098000 1068306 site-packages/torch/distributed/run.py:766] +W0621 22:11:27.098000 1068306 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:11:27.098000 1068306 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:11:27.098000 1068306 site-packages/torch/distributed/run.py:766] ***************************************** +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:11:27.246000 1991847 site-packages/torch/distributed/run.py:766] +W0621 22:11:27.246000 1991847 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:11:27.246000 1991847 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:11:27.246000 1991847 site-packages/torch/distributed/run.py:766] ***************************************** +[rank4]:[W621 22:11:50.077598097 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 4] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank2]:[W621 22:11:50.080078300 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 2] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank6]:[W621 22:11:50.081136671 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 6] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank12]:[W621 22:11:50.209845161 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 12] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank14]:[W621 22:11:50.209845286 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 14] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank10]:[W621 22:11:50.209955940 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 10] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank3]:[W621 22:11:50.086467587 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 3] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank1]:[W621 22:11:50.087142942 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 1] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank7]:[W621 22:11:50.087176269 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 7] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank5]:[W621 22:11:50.088976739 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 5] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank15]:[W621 22:11:50.223812810 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 15] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank9]:[W621 22:11:50.223864597 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 9] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank13]:[W621 22:11:50.224072711 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 13] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank11]:[W621 22:11:50.224076896 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 11] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank8]:[W621 22:11:50.312278991 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 8] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank0]:[W621 22:11:50.219477031 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 0] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +[rank0]:[W621 22:12:28.747767192 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank1]:[W621 22:12:28.932067198 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank7]:[W621 22:12:29.187715184 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank11]:[W621 22:12:29.407703757 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank15]:[W621 22:12:29.417672568 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank2]:[W621 22:12:29.351994491 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank6]:[W621 22:12:29.371739304 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank10]:[W621 22:12:29.520936166 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank14]:[W621 22:12:29.520999081 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank9]:[W621 22:12:29.649413181 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank3]:[W621 22:12:29.580791801 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank12]:[W621 22:12:29.782578954 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank13]:[W621 22:12:29.901690412 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank4]:[W621 22:12:29.855113523 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank8]:[W621 22:12:29.984743052 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank5]:[W621 22:12:30.094638269 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) ++ set +x ++ set +x ++ for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072 ++ export PROF_CTX_LENGTH=2048 ++ PROF_CTX_LENGTH=2048 ++ name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L2048*tp2.cp8.bs16.json' ++ '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L2048*tp2.cp8.bs16.json' ']' ++ echo 'Running ctx_length=2048, TP_SIZE=2, CP_SIZE=8, BATCH_SIZE=16' ++ srun bash ./attnserver.sh ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 1 --rdzv_id 343241 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-239:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 2048 --max-position-embeddings 2048 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 0 --rdzv_id 343241 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-239:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 2048 --max-position-embeddings 2048 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:12:36.452000 1995644 site-packages/torch/distributed/run.py:766] +W0621 22:12:36.452000 1995644 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:12:36.452000 1995644 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:12:36.452000 1995644 site-packages/torch/distributed/run.py:766] ***************************************** +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:12:36.618000 1072177 site-packages/torch/distributed/run.py:766] +W0621 22:12:36.618000 1072177 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:12:36.618000 1072177 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:12:36.618000 1072177 site-packages/torch/distributed/run.py:766] ***************************************** +[rank4]:[W621 22:13:00.383312673 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 4] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank2]:[W621 22:13:00.383311718 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 2] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank5]:[W621 22:13:00.383315064 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 5] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank3]:[W621 22:13:00.383374345 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 3] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank1]:[W621 22:13:00.383378245 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 1] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank13]:[W621 22:13:00.515371873 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 13] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank7]:[W621 22:13:00.383404531 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 7] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank12]:[W621 22:13:00.515686325 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 12] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank14]:[W621 22:13:00.515686089 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 14] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank10]:[W621 22:13:00.515686209 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 10] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank6]:[W621 22:13:00.383440353 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 6] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank9]:[W621 22:13:00.515727224 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 9] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank11]:[W621 22:13:00.515790739 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 11] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank15]:[W621 22:13:00.515851734 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 15] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank0]:[W621 22:13:00.521550189 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 0] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank8]:[W621 22:13:00.655489990 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 8] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +[rank1]:[W621 22:13:46.461151487 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank0]:[W621 22:13:46.518204196 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank11]:[W621 22:13:46.722148290 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank8]:[W621 22:13:46.774954874 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank3]:[W621 22:13:46.653451123 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank10]:[W621 22:13:46.855595176 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank14]:[W621 22:13:46.876055547 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank7]:[W621 22:13:46.874869657 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank9]:[W621 22:13:46.004314690 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank13]:[W621 22:13:46.064972116 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank2]:[W621 22:13:47.958091631 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank4]:[W621 22:13:47.958191026 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank5]:[W621 22:13:47.158025071 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank15]:[W621 22:13:47.368149287 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank12]:[W621 22:13:47.379423020 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank6]:[W621 22:13:47.320964635 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) ++ set +x ++ set +x ++ for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072 ++ export PROF_CTX_LENGTH=4096 ++ PROF_CTX_LENGTH=4096 ++ name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L4096*tp2.cp8.bs16.json' ++ '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L4096*tp2.cp8.bs16.json' ']' ++ echo 'Running ctx_length=4096, TP_SIZE=2, CP_SIZE=8, BATCH_SIZE=16' ++ srun bash ./attnserver.sh ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 1 --rdzv_id 343241 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-239:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 4096 --max-position-embeddings 4096 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 0 --rdzv_id 343241 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-239:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 4096 --max-position-embeddings 4096 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:13:53.622000 1075646 site-packages/torch/distributed/run.py:766] +W0621 22:13:53.622000 1075646 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:13:53.622000 1075646 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:13:53.622000 1075646 site-packages/torch/distributed/run.py:766] ***************************************** +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:13:53.636000 1999037 site-packages/torch/distributed/run.py:766] +W0621 22:13:53.636000 1999037 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:13:53.636000 1999037 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:13:53.636000 1999037 site-packages/torch/distributed/run.py:766] ***************************************** +[rank1]:[W621 22:14:17.506402869 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 1] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank5]:[W621 22:14:17.506583739 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 5] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank7]:[W621 22:14:17.506614119 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 7] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank6]:[W621 22:14:17.506715439 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 6] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank3]:[W621 22:14:17.506707684 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 3] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank2]:[W621 22:14:17.508557008 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 2] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank4]:[W621 22:14:17.509542615 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 4] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank10]:[W621 22:14:17.657940207 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 10] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank9]:[W621 22:14:17.657950759 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 9] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank13]:[W621 22:14:17.658036048 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 13] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank12]:[W621 22:14:17.658062535 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 12] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank14]:[W621 22:14:17.658082330 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 14] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank11]:[W621 22:14:17.658091798 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 11] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank15]:[W621 22:14:17.658140533 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 15] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank8]:[W621 22:14:17.742625091 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 8] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank0]:[W621 22:14:17.643336022 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 0] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +[rank1]: Traceback (most recent call last): +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank1]: pretrain( +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank1]: iteration, num_floating_point_operations_so_far = train( +[rank1]: ^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank1]: ) = train_step( +[rank1]: ^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank1]: losses_reduced = forward_backward_func( +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank1]: output_tensor, num_tokens = forward_step( +[rank1]: ^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank1]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank1]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank1]: batch = next(global_batches) +[rank1]: ^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank1]: attention_mask = torch.ones( +[rank1]: ^^^^^^^^^^^ +[rank1]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 GiB. GPU 1 has a total capacity of 139.81 GiB of which 61.04 GiB is free. Including non-PyTorch memory, this process has 78.76 GiB memory in use. Of the allocated memory 74.50 GiB is allocated by PyTorch, and 190.04 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank5]: Traceback (most recent call last): +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank5]: pretrain( +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank5]: iteration, num_floating_point_operations_so_far = train( +[rank5]: ^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank5]: ) = train_step( +[rank5]: ^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank5]: losses_reduced = forward_backward_func( +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank5]: output_tensor, num_tokens = forward_step( +[rank5]: ^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank5]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank5]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank5]: batch = next(global_batches) +[rank5]: ^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank5]: attention_mask = torch.ones( +[rank5]: ^^^^^^^^^^^ +[rank5]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 GiB. GPU 5 has a total capacity of 139.81 GiB of which 60.51 GiB is free. Including non-PyTorch memory, this process has 79.29 GiB memory in use. Of the allocated memory 74.50 GiB is allocated by PyTorch, and 190.04 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank2]: Traceback (most recent call last): +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank2]: pretrain( +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank2]: iteration, num_floating_point_operations_so_far = train( +[rank2]: ^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank2]: ) = train_step( +[rank2]: ^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank2]: losses_reduced = forward_backward_func( +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank2]: output_tensor, num_tokens = forward_step( +[rank2]: ^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank2]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank2]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank2]: batch = next(global_batches) +[rank2]: ^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank2]: attention_mask = torch.ones( +[rank2]: ^^^^^^^^^^^ +[rank2]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 GiB. GPU 2 has a total capacity of 139.81 GiB of which 59.78 GiB is free. Including non-PyTorch memory, this process has 80.02 GiB memory in use. Of the allocated memory 74.50 GiB is allocated by PyTorch, and 958.04 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank3]: Traceback (most recent call last): +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank3]: pretrain( +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank3]: iteration, num_floating_point_operations_so_far = train( +[rank3]: ^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank3]: ) = train_step( +[rank3]: ^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank3]: losses_reduced = forward_backward_func( +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank3]: output_tensor, num_tokens = forward_step( +[rank3]: ^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank3]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank3]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank3]: batch = next(global_batches) +[rank3]: ^^^^^^^^^^^^^^^^^^^^ +[rank12]: Traceback (most recent call last): +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank12]: pretrain( +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank12]: iteration, num_floating_point_operations_so_far = train( +[rank12]: ^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank12]: ) = train_step( +[rank12]: ^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank12]: losses_reduced = forward_backward_func( +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank3]: attention_mask = torch.ones( +[rank3]: ^^^^^^^^^^^ +[rank3]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 GiB. GPU 3 has a total capacity of 139.81 GiB of which 60.51 GiB is free. Including non-PyTorch memory, this process has 79.29 GiB memory in use. Of the allocated memory 74.50 GiB is allocated by PyTorch, and 190.04 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank12]: output_tensor, num_tokens = forward_step( +[rank12]: ^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank12]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank12]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank12]: batch = next(global_batches) +[rank12]: ^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank12]: attention_mask = torch.ones( +[rank12]: ^^^^^^^^^^^ +[rank12]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 GiB. GPU 4 has a total capacity of 139.81 GiB of which 60.51 GiB is free. Including non-PyTorch memory, this process has 79.29 GiB memory in use. Of the allocated memory 74.50 GiB is allocated by PyTorch, and 190.04 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank15]: Traceback (most recent call last): +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank15]: pretrain( +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank15]: iteration, num_floating_point_operations_so_far = train( +[rank15]: ^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank15]: ) = train_step( +[rank15]: ^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank15]: losses_reduced = forward_backward_func( +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank15]: output_tensor, num_tokens = forward_step( +[rank15]: ^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank15]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank15]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank15]: batch = next(global_batches) +[rank15]: ^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank15]: attention_mask = torch.ones( +[rank15]: ^^^^^^^^^^^ +[rank15]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 GiB. GPU 7 has a total capacity of 139.81 GiB of which 61.05 GiB is free. Including non-PyTorch memory, this process has 78.75 GiB memory in use. Of the allocated memory 74.50 GiB is allocated by PyTorch, and 190.04 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank7]: Traceback (most recent call last): +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank7]: pretrain( +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank7]: iteration, num_floating_point_operations_so_far = train( +[rank7]: ^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank7]: ) = train_step( +[rank7]: ^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank7]: losses_reduced = forward_backward_func( +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank7]: output_tensor, num_tokens = forward_step( +[rank7]: ^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank7]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank7]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank7]: batch = next(global_batches) +[rank7]: ^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank7]: attention_mask = torch.ones( +[rank7]: ^^^^^^^^^^^ +[rank7]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 GiB. GPU 7 has a total capacity of 139.81 GiB of which 61.04 GiB is free. Including non-PyTorch memory, this process has 78.76 GiB memory in use. Of the allocated memory 74.50 GiB is allocated by PyTorch, and 190.04 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank0]: Traceback (most recent call last): +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank0]: pretrain( +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank0]: iteration, num_floating_point_operations_so_far = train( +[rank0]: ^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank0]: ) = train_step( +[rank0]: ^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank0]: losses_reduced = forward_backward_func( +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank0]: output_tensor, num_tokens = forward_step( +[rank0]: ^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank0]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank0]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank0]: batch = next(global_batches) +[rank0]: ^^^^^^^^^^^^^^^^^^^^ +[rank14]: Traceback (most recent call last): +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank14]: pretrain( +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank14]: iteration, num_floating_point_operations_so_far = train( +[rank14]: ^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank14]: ) = train_step( +[rank14]: ^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank14]: losses_reduced = forward_backward_func( +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank0]: attention_mask = torch.ones( +[rank0]: ^^^^^^^^^^^ +[rank0]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 GiB. GPU 0 has a total capacity of 139.81 GiB of which 61.05 GiB is free. Including non-PyTorch memory, this process has 78.75 GiB memory in use. Of the allocated memory 74.50 GiB is allocated by PyTorch, and 190.04 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank14]: output_tensor, num_tokens = forward_step( +[rank14]: ^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank14]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank14]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank14]: batch = next(global_batches) +[rank14]: ^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank14]: attention_mask = torch.ones( +[rank14]: ^^^^^^^^^^^ +[rank14]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 GiB. GPU 6 has a total capacity of 139.81 GiB of which 61.04 GiB is free. Including non-PyTorch memory, this process has 78.76 GiB memory in use. Of the allocated memory 74.50 GiB is allocated by PyTorch, and 190.04 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank9]: Traceback (most recent call last): +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank9]: pretrain( +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank9]: iteration, num_floating_point_operations_so_far = train( +[rank9]: ^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank9]: ) = train_step( +[rank9]: ^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank9]: losses_reduced = forward_backward_func( +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank9]: output_tensor, num_tokens = forward_step( +[rank9]: ^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank9]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank9]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank9]: batch = next(global_batches) +[rank9]: ^^^^^^^^^^^^^^^^^^^^ +[rank6]: Traceback (most recent call last): +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank6]: pretrain( +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank6]: iteration, num_floating_point_operations_so_far = train( +[rank6]: ^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank6]: ) = train_step( +[rank6]: ^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank6]: losses_reduced = forward_backward_func( +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank9]: attention_mask = torch.ones( +[rank9]: ^^^^^^^^^^^ +[rank9]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 GiB. GPU 1 has a total capacity of 139.81 GiB of which 61.05 GiB is free. Including non-PyTorch memory, this process has 78.75 GiB memory in use. Of the allocated memory 74.50 GiB is allocated by PyTorch, and 190.04 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank6]: output_tensor, num_tokens = forward_step( +[rank6]: ^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank6]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank6]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank6]: batch = next(global_batches) +[rank6]: ^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank6]: attention_mask = torch.ones( +[rank6]: ^^^^^^^^^^^ +[rank6]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 GiB. GPU 6 has a total capacity of 139.81 GiB of which 61.05 GiB is free. Including non-PyTorch memory, this process has 78.75 GiB memory in use. Of the allocated memory 74.50 GiB is allocated by PyTorch, and 190.04 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank4]: Traceback (most recent call last): +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank4]: pretrain( +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank4]: iteration, num_floating_point_operations_so_far = train( +[rank4]: ^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank4]: ) = train_step( +[rank4]: ^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank4]: losses_reduced = forward_backward_func( +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank4]: output_tensor, num_tokens = forward_step( +[rank4]: ^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank4]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank4]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank4]: batch = next(global_batches) +[rank4]: ^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank4]: attention_mask = torch.ones( +[rank4]: ^^^^^^^^^^^ +[rank4]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 GiB. GPU 4 has a total capacity of 139.81 GiB of which 60.53 GiB is free. Including non-PyTorch memory, this process has 79.27 GiB memory in use. Of the allocated memory 74.50 GiB is allocated by PyTorch, and 190.04 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank11]: Traceback (most recent call last): +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank11]: pretrain( +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank11]: iteration, num_floating_point_operations_so_far = train( +[rank11]: ^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank11]: ) = train_step( +[rank11]: ^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank11]: losses_reduced = forward_backward_func( +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank11]: output_tensor, num_tokens = forward_step( +[rank11]: ^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank11]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank11]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank11]: batch = next(global_batches) +[rank11]: ^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank11]: attention_mask = torch.ones( +[rank11]: ^^^^^^^^^^^ +[rank11]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 GiB. GPU 3 has a total capacity of 139.81 GiB of which 60.51 GiB is free. Including non-PyTorch memory, this process has 79.29 GiB memory in use. Of the allocated memory 74.50 GiB is allocated by PyTorch, and 206.04 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank10]: Traceback (most recent call last): +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank10]: pretrain( +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank10]: iteration, num_floating_point_operations_so_far = train( +[rank10]: ^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank10]: ) = train_step( +[rank10]: ^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank10]: losses_reduced = forward_backward_func( +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank10]: output_tensor, num_tokens = forward_step( +[rank10]: ^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank10]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank10]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank10]: batch = next(global_batches) +[rank10]: ^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank10]: attention_mask = torch.ones( +[rank10]: ^^^^^^^^^^^ +[rank10]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 GiB. GPU 2 has a total capacity of 139.81 GiB of which 60.50 GiB is free. Including non-PyTorch memory, this process has 79.30 GiB memory in use. Of the allocated memory 74.50 GiB is allocated by PyTorch, and 206.04 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank8]: Traceback (most recent call last): +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank8]: pretrain( +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank8]: iteration, num_floating_point_operations_so_far = train( +[rank8]: ^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank8]: ) = train_step( +[rank8]: ^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank8]: losses_reduced = forward_backward_func( +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank8]: output_tensor, num_tokens = forward_step( +[rank8]: ^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank8]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank8]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank8]: batch = next(global_batches) +[rank8]: ^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank8]: attention_mask = torch.ones( +[rank8]: ^^^^^^^^^^^ +[rank8]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 GiB. GPU 0 has a total capacity of 139.81 GiB of which 61.04 GiB is free. Including non-PyTorch memory, this process has 78.76 GiB memory in use. Of the allocated memory 74.50 GiB is allocated by PyTorch, and 190.04 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank13]: Traceback (most recent call last): +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank13]: pretrain( +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank13]: iteration, num_floating_point_operations_so_far = train( +[rank13]: ^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank13]: ) = train_step( +[rank13]: ^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank13]: losses_reduced = forward_backward_func( +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank13]: output_tensor, num_tokens = forward_step( +[rank13]: ^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank13]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank13]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank13]: batch = next(global_batches) +[rank13]: ^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank13]: attention_mask = torch.ones( +[rank13]: ^^^^^^^^^^^ +[rank13]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 GiB. GPU 5 has a total capacity of 139.81 GiB of which 60.53 GiB is free. Including non-PyTorch memory, this process has 79.27 GiB memory in use. Of the allocated memory 74.50 GiB is allocated by PyTorch, and 190.04 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank1]:[W621 22:14:45.523137468 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank3]:[W621 22:14:46.960053700 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank7]:[W621 22:14:46.969913785 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank9]:[W621 22:14:46.230140183 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank11]:[W621 22:14:46.260506728 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank15]:[W621 22:14:46.270502875 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank5]:[W621 22:14:46.272327385 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank13]:[W621 22:14:46.603208421 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +W0621 22:14:48.173000 1075646 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1075734 closing signal SIGTERM +W0621 22:14:48.175000 1075646 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1075736 closing signal SIGTERM +W0621 22:14:48.178000 1075646 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1075737 closing signal SIGTERM +W0621 22:14:48.180000 1075646 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1075738 closing signal SIGTERM +W0621 22:14:48.184000 1075646 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1075739 closing signal SIGTERM +W0621 22:14:48.185000 1075646 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1075740 closing signal SIGTERM +W0621 22:14:48.187000 1075646 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1075741 closing signal SIGTERM +W0621 22:14:48.279000 1999037 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1999109 closing signal SIGTERM +W0621 22:14:48.282000 1999037 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1999110 closing signal SIGTERM +W0621 22:14:48.283000 1999037 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1999111 closing signal SIGTERM +W0621 22:14:48.285000 1999037 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1999113 closing signal SIGTERM +W0621 22:14:48.289000 1999037 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1999114 closing signal SIGTERM +W0621 22:14:48.290000 1999037 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1999115 closing signal SIGTERM +W0621 22:14:48.298000 1999037 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1999116 closing signal SIGTERM +E0621 22:14:50.286000 1999037 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 3 (pid: 1999112) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +Traceback (most recent call last): + File "", line 198, in _run_module_as_main + File "", line 88, in _run_code + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in + main() + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper + return arg(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main + launch(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch + run(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run + elastic_launch( + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__ + return launch_agent(self._config, self._entrypoint, list(args)) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent + raise ChildFailedError( +torch.distributed.elastic.multiprocessing.errors.ChildFailedError: +============================================================ +./pretrain_gpt_profile.py FAILED +------------------------------------------------------------ +Failures: + +------------------------------------------------------------ +Root Cause (first observed failure): +[0]: + time : 2025-06-21_22:14:48 + host : fs-mbz-gpu-286 + rank : 11 (local_rank: 3) + exitcode : 1 (pid: 1999112) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +============================================================ +E0621 22:14:50.454000 1075646 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 1 (pid: 1075735) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +Traceback (most recent call last): + File "", line 198, in _run_module_as_main + File "", line 88, in _run_code + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in + main() + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper + return arg(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main + launch(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch + run(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run + elastic_launch( + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__ + return launch_agent(self._config, self._entrypoint, list(args)) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent + raise ChildFailedError( +torch.distributed.elastic.multiprocessing.errors.ChildFailedError: +============================================================ +./pretrain_gpt_profile.py FAILED +------------------------------------------------------------ +Failures: + +------------------------------------------------------------ +Root Cause (first observed failure): +[0]: + time : 2025-06-21_22:14:48 + host : fs-mbz-gpu-239 + rank : 1 (local_rank: 1) + exitcode : 1 (pid: 1075735) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +============================================================ ++ set +x ++ set +x ++ for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072 ++ export PROF_CTX_LENGTH=8192 ++ PROF_CTX_LENGTH=8192 ++ name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L8192*tp2.cp8.bs16.json' ++ '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L8192*tp2.cp8.bs16.json' ']' ++ echo 'Running ctx_length=8192, TP_SIZE=2, CP_SIZE=8, BATCH_SIZE=16' ++ srun bash ./attnserver.sh ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 0 --rdzv_id 343241 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-239:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 8192 --max-position-embeddings 8192 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 1 --rdzv_id 343241 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-239:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 8192 --max-position-embeddings 8192 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:14:53.639000 2001972 site-packages/torch/distributed/run.py:766] +W0621 22:14:53.639000 2001972 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:14:53.639000 2001972 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:14:53.639000 2001972 site-packages/torch/distributed/run.py:766] ***************************************** +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:14:53.651000 1078649 site-packages/torch/distributed/run.py:766] +W0621 22:14:53.651000 1078649 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:14:53.651000 1078649 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:14:53.651000 1078649 site-packages/torch/distributed/run.py:766] ***************************************** +[rank6]:[W621 22:15:17.346116659 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 6] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank4]:[W621 22:15:17.346142258 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 4] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank2]:[W621 22:15:17.346206351 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 2] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank5]:[W621 22:15:17.354351738 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 5] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank7]:[W621 22:15:17.354387499 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 7] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank3]:[W621 22:15:17.354456951 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 3] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank1]:[W621 22:15:17.354531849 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 1] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank14]:[W621 22:15:17.487335094 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 14] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank15]:[W621 22:15:17.487725154 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 15] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank10]:[W621 22:15:17.487776814 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 10] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank11]:[W621 22:15:17.487843268 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 11] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank9]:[W621 22:15:17.487902912 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 9] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank12]:[W621 22:15:17.487982921 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 12] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank13]:[W621 22:15:17.488581035 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 13] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank8]:[W621 22:15:17.569170293 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 8] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank0]:[W621 22:15:17.485718935 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 0] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +[rank15]: Traceback (most recent call last): +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank15]: pretrain( +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank15]: iteration, num_floating_point_operations_so_far = train( +[rank15]: ^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank15]: ) = train_step( +[rank15]: ^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank15]: losses_reduced = forward_backward_func( +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank15]: output_tensor, num_tokens = forward_step( +[rank15]: ^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank15]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank15]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank15]: batch = next(global_batches) +[rank15]: ^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank15]: attention_mask = torch.ones( +[rank15]: ^^^^^^^^^^^ +[rank15]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 256.00 GiB. GPU 7 has a total capacity of 139.81 GiB of which 135.89 GiB is free. Including non-PyTorch memory, this process has 3.91 GiB memory in use. Of the allocated memory 2.37 GiB is allocated by PyTorch, and 63.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank8]: Traceback (most recent call last): +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank8]: pretrain( +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank8]: iteration, num_floating_point_operations_so_far = train( +[rank8]: ^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank8]: ) = train_step( +[rank8]: ^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank8]: losses_reduced = forward_backward_func( +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank8]: output_tensor, num_tokens = forward_step( +[rank8]: ^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank8]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank8]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank8]: batch = next(global_batches) +[rank8]: ^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank8]: attention_mask = torch.ones( +[rank8]: ^^^^^^^^^^^ +[rank8]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 256.00 GiB. GPU 0 has a total capacity of 139.81 GiB of which 135.87 GiB is free. Including non-PyTorch memory, this process has 3.93 GiB memory in use. Of the allocated memory 2.37 GiB is allocated by PyTorch, and 63.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank7]: Traceback (most recent call last): +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank7]: pretrain( +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank7]: iteration, num_floating_point_operations_so_far = train( +[rank7]: ^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank7]: ) = train_step( +[rank7]: ^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank7]: losses_reduced = forward_backward_func( +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank7]: output_tensor, num_tokens = forward_step( +[rank7]: ^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank7]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank7]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank7]: batch = next(global_batches) +[rank7]: ^^^^^^^^^^^^^^^^^^^^ +[rank9]: Traceback (most recent call last): +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank9]: pretrain( +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank9]: iteration, num_floating_point_operations_so_far = train( +[rank9]: ^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank9]: ) = train_step( +[rank9]: ^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank9]: losses_reduced = forward_backward_func( +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank7]: attention_mask = torch.ones( +[rank7]: ^^^^^^^^^^^ +[rank7]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 256.00 GiB. GPU 7 has a total capacity of 139.81 GiB of which 135.87 GiB is free. Including non-PyTorch memory, this process has 3.93 GiB memory in use. Of the allocated memory 2.37 GiB is allocated by PyTorch, and 63.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank5]: Traceback (most recent call last): +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank5]: pretrain( +[rank9]: output_tensor, num_tokens = forward_step( +[rank9]: ^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank9]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank9]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank9]: batch = next(global_batches) +[rank9]: ^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank5]: iteration, num_floating_point_operations_so_far = train( +[rank5]: ^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank5]: ) = train_step( +[rank5]: ^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank5]: losses_reduced = forward_backward_func( +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank5]: output_tensor, num_tokens = forward_step( +[rank5]: ^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank9]: attention_mask = torch.ones( +[rank9]: ^^^^^^^^^^^ +[rank9]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 256.00 GiB. GPU 1 has a total capacity of 139.81 GiB of which 135.89 GiB is free. Including non-PyTorch memory, this process has 3.91 GiB memory in use. Of the allocated memory 2.37 GiB is allocated by PyTorch, and 63.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank13]: Traceback (most recent call last): +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank13]: pretrain( +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank13]: iteration, num_floating_point_operations_so_far = train( +[rank13]: ^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank13]: ) = train_step( +[rank13]: ^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank13]: losses_reduced = forward_backward_func( +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank5]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank5]: batch = next(global_batches) +[rank5]: ^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank5]: attention_mask = torch.ones( +[rank5]: ^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank13]: output_tensor, num_tokens = forward_step( +[rank13]: ^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank13]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank13]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 256.00 GiB. GPU 5 has a total capacity of 139.81 GiB of which 135.87 GiB is free. Including non-PyTorch memory, this process has 3.93 GiB memory in use. Of the allocated memory 2.37 GiB is allocated by PyTorch, and 63.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank1]: Traceback (most recent call last): +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank1]: pretrain( +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank1]: iteration, num_floating_point_operations_so_far = train( +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank13]: batch = next(global_batches) +[rank13]: ^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank13]: attention_mask = torch.ones( +[rank13]: ^^^^^^^^^^^ +[rank13]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 256.00 GiB. GPU 5 has a total capacity of 139.81 GiB of which 135.89 GiB is free. Including non-PyTorch memory, this process has 3.91 GiB memory in use. Of the allocated memory 2.37 GiB is allocated by PyTorch, and 63.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank1]: ^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank1]: ) = train_step( +[rank1]: ^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank1]: losses_reduced = forward_backward_func( +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank1]: output_tensor, num_tokens = forward_step( +[rank1]: ^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank1]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank11]: Traceback (most recent call last): +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank11]: pretrain( +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank11]: iteration, num_floating_point_operations_so_far = train( +[rank11]: ^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank11]: ) = train_step( +[rank11]: ^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank11]: losses_reduced = forward_backward_func( +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank1]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank1]: batch = next(global_batches) +[rank1]: ^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank1]: attention_mask = torch.ones( +[rank1]: ^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank11]: output_tensor, num_tokens = forward_step( +[rank11]: ^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank11]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank11]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 256.00 GiB. GPU 1 has a total capacity of 139.81 GiB of which 135.87 GiB is free. Including non-PyTorch memory, this process has 3.93 GiB memory in use. Of the allocated memory 2.37 GiB is allocated by PyTorch, and 63.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank11]: batch = next(global_batches) +[rank11]: ^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank11]: attention_mask = torch.ones( +[rank11]: ^^^^^^^^^^^ +[rank11]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 256.00 GiB. GPU 3 has a total capacity of 139.81 GiB of which 135.89 GiB is free. Including non-PyTorch memory, this process has 3.91 GiB memory in use. Of the allocated memory 2.37 GiB is allocated by PyTorch, and 63.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank6]: Traceback (most recent call last): +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank6]: pretrain( +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank6]: iteration, num_floating_point_operations_so_far = train( +[rank6]: ^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank6]: ) = train_step( +[rank6]: ^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank6]: losses_reduced = forward_backward_func( +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank6]: output_tensor, num_tokens = forward_step( +[rank6]: ^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank6]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank6]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank6]: batch = next(global_batches) +[rank6]: ^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank6]: attention_mask = torch.ones( +[rank6]: ^^^^^^^^^^^ +[rank6]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 256.00 GiB. GPU 6 has a total capacity of 139.81 GiB of which 135.89 GiB is free. Including non-PyTorch memory, this process has 3.91 GiB memory in use. Of the allocated memory 2.37 GiB is allocated by PyTorch, and 63.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank3]: Traceback (most recent call last): +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank3]: pretrain( +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank3]: iteration, num_floating_point_operations_so_far = train( +[rank3]: ^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank3]: ) = train_step( +[rank3]: ^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank3]: losses_reduced = forward_backward_func( +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank12]: Traceback (most recent call last): +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank12]: pretrain( +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank12]: iteration, num_floating_point_operations_so_far = train( +[rank12]: ^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank12]: ) = train_step( +[rank12]: ^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank12]: losses_reduced = forward_backward_func( +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: output_tensor, num_tokens = forward_step( +[rank3]: ^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank3]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank3]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank3]: batch = next(global_batches) +[rank3]: ^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank12]: output_tensor, num_tokens = forward_step( +[rank12]: ^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank12]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank12]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank3]: attention_mask = torch.ones( +[rank3]: ^^^^^^^^^^^ +[rank3]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 256.00 GiB. GPU 3 has a total capacity of 139.81 GiB of which 135.87 GiB is free. Including non-PyTorch memory, this process has 3.93 GiB memory in use. Of the allocated memory 2.37 GiB is allocated by PyTorch, and 63.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank12]: batch = next(global_batches) +[rank12]: ^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank12]: attention_mask = torch.ones( +[rank12]: ^^^^^^^^^^^ +[rank12]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 256.00 GiB. GPU 4 has a total capacity of 139.81 GiB of which 135.87 GiB is free. Including non-PyTorch memory, this process has 3.93 GiB memory in use. Of the allocated memory 2.37 GiB is allocated by PyTorch, and 63.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank0]: Traceback (most recent call last): +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank0]: pretrain( +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank0]: iteration, num_floating_point_operations_so_far = train( +[rank0]: ^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank0]: ) = train_step( +[rank0]: ^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank0]: losses_reduced = forward_backward_func( +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank10]: Traceback (most recent call last): +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank10]: pretrain( +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank10]: iteration, num_floating_point_operations_so_far = train( +[rank10]: ^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank10]: ) = train_step( +[rank10]: ^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank10]: losses_reduced = forward_backward_func( +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: output_tensor, num_tokens = forward_step( +[rank0]: ^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank0]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank0]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank0]: batch = next(global_batches) +[rank0]: ^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank10]: output_tensor, num_tokens = forward_step( +[rank10]: ^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank10]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank10]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank0]: attention_mask = torch.ones( +[rank0]: ^^^^^^^^^^^ +[rank0]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 256.00 GiB. GPU 0 has a total capacity of 139.81 GiB of which 135.89 GiB is free. Including non-PyTorch memory, this process has 3.91 GiB memory in use. Of the allocated memory 2.37 GiB is allocated by PyTorch, and 63.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank10]: batch = next(global_batches) +[rank10]: ^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank10]: attention_mask = torch.ones( +[rank10]: ^^^^^^^^^^^ +[rank10]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 256.00 GiB. GPU 2 has a total capacity of 139.81 GiB of which 135.87 GiB is free. Including non-PyTorch memory, this process has 3.93 GiB memory in use. Of the allocated memory 2.37 GiB is allocated by PyTorch, and 63.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank4]: Traceback (most recent call last): +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank4]: pretrain( +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank4]: iteration, num_floating_point_operations_so_far = train( +[rank4]: ^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank4]: ) = train_step( +[rank4]: ^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank4]: losses_reduced = forward_backward_func( +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank14]: Traceback (most recent call last): +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank14]: pretrain( +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank14]: iteration, num_floating_point_operations_so_far = train( +[rank14]: ^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank14]: ) = train_step( +[rank14]: ^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank14]: losses_reduced = forward_backward_func( +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: output_tensor, num_tokens = forward_step( +[rank4]: ^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank4]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank4]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank4]: batch = next(global_batches) +[rank4]: ^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank14]: output_tensor, num_tokens = forward_step( +[rank14]: ^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank14]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank14]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank4]: attention_mask = torch.ones( +[rank4]: ^^^^^^^^^^^ +[rank4]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 256.00 GiB. GPU 4 has a total capacity of 139.81 GiB of which 135.89 GiB is free. Including non-PyTorch memory, this process has 3.91 GiB memory in use. Of the allocated memory 2.37 GiB is allocated by PyTorch, and 63.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank14]: batch = next(global_batches) +[rank14]: ^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank14]: attention_mask = torch.ones( +[rank14]: ^^^^^^^^^^^ +[rank14]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 256.00 GiB. GPU 6 has a total capacity of 139.81 GiB of which 135.87 GiB is free. Including non-PyTorch memory, this process has 3.93 GiB memory in use. Of the allocated memory 2.37 GiB is allocated by PyTorch, and 63.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank2]: Traceback (most recent call last): +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank2]: pretrain( +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank2]: iteration, num_floating_point_operations_so_far = train( +[rank2]: ^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank2]: ) = train_step( +[rank2]: ^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank2]: losses_reduced = forward_backward_func( +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank2]: output_tensor, num_tokens = forward_step( +[rank2]: ^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank2]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank2]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank2]: batch = next(global_batches) +[rank2]: ^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank2]: attention_mask = torch.ones( +[rank2]: ^^^^^^^^^^^ +[rank2]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 256.00 GiB. GPU 2 has a total capacity of 139.81 GiB of which 135.89 GiB is free. Including non-PyTorch memory, this process has 3.91 GiB memory in use. Of the allocated memory 2.37 GiB is allocated by PyTorch, and 63.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank5]:[W621 22:15:27.120150973 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank15]:[W621 22:15:27.420848572 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank9]:[W621 22:15:27.421149946 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank13]:[W621 22:15:27.431159953 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank1]:[W621 22:15:27.308045002 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank11]:[W621 22:15:27.471420541 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank3]:[W621 22:15:27.342275403 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank7]:[W621 22:15:27.352284244 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +W0621 22:15:28.475000 1078649 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1078720 closing signal SIGTERM +W0621 22:15:28.478000 1078649 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1078721 closing signal SIGTERM +W0621 22:15:28.478000 1078649 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1078722 closing signal SIGTERM +W0621 22:15:28.481000 1078649 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1078723 closing signal SIGTERM +W0621 22:15:28.482000 1078649 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1078724 closing signal SIGTERM +W0621 22:15:28.485000 1078649 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1078726 closing signal SIGTERM +W0621 22:15:28.500000 1078649 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1078727 closing signal SIGTERM +W0621 22:15:28.717000 2001972 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2002057 closing signal SIGTERM +W0621 22:15:28.720000 2001972 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2002059 closing signal SIGTERM +W0621 22:15:28.723000 2001972 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2002060 closing signal SIGTERM +W0621 22:15:28.724000 2001972 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2002061 closing signal SIGTERM +W0621 22:15:28.727000 2001972 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2002062 closing signal SIGTERM +W0621 22:15:28.728000 2001972 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2002063 closing signal SIGTERM +W0621 22:15:28.730000 2001972 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2002064 closing signal SIGTERM +E0621 22:15:29.020000 2001972 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 1 (pid: 2002058) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +Traceback (most recent call last): + File "", line 198, in _run_module_as_main + File "", line 88, in _run_code + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in + main() + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper + return arg(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main + launch(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch + run(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run + elastic_launch( + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__ + return launch_agent(self._config, self._entrypoint, list(args)) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent + raise ChildFailedError( +torch.distributed.elastic.multiprocessing.errors.ChildFailedError: +============================================================ +./pretrain_gpt_profile.py FAILED +------------------------------------------------------------ +Failures: + +------------------------------------------------------------ +Root Cause (first observed failure): +[0]: + time : 2025-06-21_22:15:28 + host : fs-mbz-gpu-286 + rank : 9 (local_rank: 1) + exitcode : 1 (pid: 2002058) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +============================================================ ++ set +x +E0621 22:15:29.722000 1078649 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 5 (pid: 1078725) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +Traceback (most recent call last): + File "", line 198, in _run_module_as_main + File "", line 88, in _run_code + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in + main() + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper + return arg(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main + launch(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch + run(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run + elastic_launch( + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__ + return launch_agent(self._config, self._entrypoint, list(args)) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent + raise ChildFailedError( +torch.distributed.elastic.multiprocessing.errors.ChildFailedError: +============================================================ +./pretrain_gpt_profile.py FAILED +------------------------------------------------------------ +Failures: + +------------------------------------------------------------ +Root Cause (first observed failure): +[0]: + time : 2025-06-21_22:15:28 + host : fs-mbz-gpu-239 + rank : 5 (local_rank: 5) + exitcode : 1 (pid: 1078725) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +============================================================ ++ set +x ++ for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072 ++ export PROF_CTX_LENGTH=12288 ++ PROF_CTX_LENGTH=12288 ++ name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L12288*tp2.cp8.bs16.json' ++ '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L12288*tp2.cp8.bs16.json' ']' ++ echo 'Running ctx_length=12288, TP_SIZE=2, CP_SIZE=8, BATCH_SIZE=16' ++ srun bash ./attnserver.sh ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 0 --rdzv_id 343241 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-239:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 12288 --max-position-embeddings 12288 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 1 --rdzv_id 343241 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-239:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 12288 --max-position-embeddings 12288 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:15:33.045000 1080631 site-packages/torch/distributed/run.py:766] +W0621 22:15:33.045000 1080631 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:15:33.045000 1080631 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:15:33.045000 1080631 site-packages/torch/distributed/run.py:766] ***************************************** +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:15:33.073000 2003918 site-packages/torch/distributed/run.py:766] +W0621 22:15:33.073000 2003918 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:15:33.073000 2003918 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:15:33.073000 2003918 site-packages/torch/distributed/run.py:766] ***************************************** +[rank4]:[W621 22:15:55.675985357 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 4] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank2]:[W621 22:15:55.675987203 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 2] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank5]:[W621 22:15:55.676571306 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 5] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank3]:[W621 22:15:55.677175710 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 3] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank9]:[W621 22:15:55.812783132 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 9] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank13]:[W621 22:15:55.812799415 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 13] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank15]:[W621 22:15:55.812796551 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 15] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank7]:[W621 22:15:55.677258897 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 7] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank10]:[W621 22:15:55.812798878 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 10] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank1]:[W621 22:15:55.680729420 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 1] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank12]:[W621 22:15:55.812905727 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 12] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank14]:[W621 22:15:55.812908106 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 14] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank11]:[W621 22:15:55.812989427 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 11] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank6]:[W621 22:15:55.683612441 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 6] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank8]:[W621 22:15:55.897800255 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 8] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank0]:[W621 22:15:55.820796653 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 0] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +[rank9]: Traceback (most recent call last): +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank9]: pretrain( +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank9]: iteration, num_floating_point_operations_so_far = train( +[rank9]: ^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank9]: ) = train_step( +[rank9]: ^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank9]: losses_reduced = forward_backward_func( +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank9]: output_tensor, num_tokens = forward_step( +[rank9]: ^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank9]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank9]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank9]: batch = next(global_batches) +[rank9]: ^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank9]: attention_mask = torch.ones( +[rank9]: ^^^^^^^^^^^ +[rank9]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 576.00 GiB. GPU 1 has a total capacity of 139.81 GiB of which 135.72 GiB is free. Including non-PyTorch memory, this process has 4.09 GiB memory in use. Of the allocated memory 2.52 GiB is allocated by PyTorch, and 91.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank1]: Traceback (most recent call last): +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank1]: pretrain( +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank1]: iteration, num_floating_point_operations_so_far = train( +[rank1]: ^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank1]: ) = train_step( +[rank1]: ^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank1]: losses_reduced = forward_backward_func( +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank1]: output_tensor, num_tokens = forward_step( +[rank1]: ^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank1]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank1]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank1]: batch = next(global_batches) +[rank1]: ^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank1]: attention_mask = torch.ones( +[rank1]: ^^^^^^^^^^^ +[rank1]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 576.00 GiB. GPU 1 has a total capacity of 139.81 GiB of which 135.70 GiB is free. Including non-PyTorch memory, this process has 4.10 GiB memory in use. Of the allocated memory 2.52 GiB is allocated by PyTorch, and 91.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank2]: Traceback (most recent call last): +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank2]: pretrain( +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank2]: iteration, num_floating_point_operations_so_far = train( +[rank2]: ^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank2]: ) = train_step( +[rank2]: ^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank2]: losses_reduced = forward_backward_func( +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank2]: output_tensor, num_tokens = forward_step( +[rank2]: ^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank2]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank2]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank2]: batch = next(global_batches) +[rank2]: ^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank2]: attention_mask = torch.ones( +[rank2]: ^^^^^^^^^^^ +[rank2]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 576.00 GiB. GPU 2 has a total capacity of 139.81 GiB of which 135.72 GiB is free. Including non-PyTorch memory, this process has 4.09 GiB memory in use. Of the allocated memory 2.52 GiB is allocated by PyTorch, and 91.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank3]: Traceback (most recent call last): +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank3]: pretrain( +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank3]: iteration, num_floating_point_operations_so_far = train( +[rank3]: ^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank3]: ) = train_step( +[rank3]: ^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank3]: losses_reduced = forward_backward_func( +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank3]: output_tensor, num_tokens = forward_step( +[rank3]: ^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank3]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank3]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank3]: batch = next(global_batches) +[rank3]: ^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank3]: attention_mask = torch.ones( +[rank3]: ^^^^^^^^^^^ +[rank15]: Traceback (most recent call last): +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank15]: pretrain( +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank15]: iteration, num_floating_point_operations_so_far = train( +[rank15]: ^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank15]: ) = train_step( +[rank15]: ^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank15]: losses_reduced = forward_backward_func( +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 576.00 GiB. GPU 3 has a total capacity of 139.81 GiB of which 135.70 GiB is free. Including non-PyTorch memory, this process has 4.10 GiB memory in use. Of the allocated memory 2.52 GiB is allocated by PyTorch, and 91.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank15]: output_tensor, num_tokens = forward_step( +[rank15]: ^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank15]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank15]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank15]: batch = next(global_batches) +[rank15]: ^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank15]: attention_mask = torch.ones( +[rank15]: ^^^^^^^^^^^ +[rank15]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 576.00 GiB. GPU 7 has a total capacity of 139.81 GiB of which 135.72 GiB is free. Including non-PyTorch memory, this process has 4.09 GiB memory in use. Of the allocated memory 2.52 GiB is allocated by PyTorch, and 91.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank13]: Traceback (most recent call last): +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank13]: pretrain( +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank13]: iteration, num_floating_point_operations_so_far = train( +[rank13]: ^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank13]: ) = train_step( +[rank13]: ^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank13]: losses_reduced = forward_backward_func( +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank13]: output_tensor, num_tokens = forward_step( +[rank13]: ^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank13]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank13]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: Traceback (most recent call last): +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank4]: pretrain( +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank4]: iteration, num_floating_point_operations_so_far = train( +[rank4]: ^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank4]: ) = train_step( +[rank4]: ^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank4]: losses_reduced = forward_backward_func( +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank13]: batch = next(global_batches) +[rank13]: ^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank13]: attention_mask = torch.ones( +[rank13]: ^^^^^^^^^^^ +[rank13]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 576.00 GiB. GPU 5 has a total capacity of 139.81 GiB of which 135.72 GiB is free. Including non-PyTorch memory, this process has 4.09 GiB memory in use. Of the allocated memory 2.52 GiB is allocated by PyTorch, and 91.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank4]: output_tensor, num_tokens = forward_step( +[rank4]: ^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank4]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank4]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank4]: batch = next(global_batches) +[rank4]: ^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank4]: attention_mask = torch.ones( +[rank4]: ^^^^^^^^^^^ +[rank4]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 576.00 GiB. GPU 4 has a total capacity of 139.81 GiB of which 135.72 GiB is free. Including non-PyTorch memory, this process has 4.09 GiB memory in use. Of the allocated memory 2.52 GiB is allocated by PyTorch, and 91.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank5]: Traceback (most recent call last): +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank5]: pretrain( +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank5]: iteration, num_floating_point_operations_so_far = train( +[rank5]: ^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank5]: ) = train_step( +[rank5]: ^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank5]: losses_reduced = forward_backward_func( +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank5]: output_tensor, num_tokens = forward_step( +[rank5]: ^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank5]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank5]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank5]: batch = next(global_batches) +[rank5]: ^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank5]: attention_mask = torch.ones( +[rank5]: ^^^^^^^^^^^ +[rank5]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 576.00 GiB. GPU 5 has a total capacity of 139.81 GiB of which 135.70 GiB is free. Including non-PyTorch memory, this process has 4.10 GiB memory in use. Of the allocated memory 2.52 GiB is allocated by PyTorch, and 91.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank7]: Traceback (most recent call last): +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank7]: pretrain( +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank7]: iteration, num_floating_point_operations_so_far = train( +[rank7]: ^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank7]: ) = train_step( +[rank7]: ^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank7]: losses_reduced = forward_backward_func( +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank8]: Traceback (most recent call last): +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank8]: pretrain( +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank8]: iteration, num_floating_point_operations_so_far = train( +[rank8]: ^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank8]: ) = train_step( +[rank8]: ^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank8]: losses_reduced = forward_backward_func( +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank7]: output_tensor, num_tokens = forward_step( +[rank7]: ^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank7]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank7]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank7]: batch = next(global_batches) +[rank7]: ^^^^^^^^^^^^^^^^^^^^ +[rank8]: output_tensor, num_tokens = forward_step( +[rank8]: ^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank8]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank8]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank8]: batch = next(global_batches) +[rank8]: ^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank7]: attention_mask = torch.ones( +[rank7]: ^^^^^^^^^^^ +[rank7]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 576.00 GiB. GPU 7 has a total capacity of 139.81 GiB of which 135.70 GiB is free. Including non-PyTorch memory, this process has 4.10 GiB memory in use. Of the allocated memory 2.52 GiB is allocated by PyTorch, and 91.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank8]: attention_mask = torch.ones( +[rank8]: ^^^^^^^^^^^ +[rank8]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 576.00 GiB. GPU 0 has a total capacity of 139.81 GiB of which 135.70 GiB is free. Including non-PyTorch memory, this process has 4.10 GiB memory in use. Of the allocated memory 2.52 GiB is allocated by PyTorch, and 91.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank0]: Traceback (most recent call last): +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank0]: pretrain( +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank0]: iteration, num_floating_point_operations_so_far = train( +[rank0]: ^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank0]: ) = train_step( +[rank0]: ^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank0]: losses_reduced = forward_backward_func( +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank11]: Traceback (most recent call last): +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank11]: pretrain( +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank11]: iteration, num_floating_point_operations_so_far = train( +[rank11]: ^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank11]: ) = train_step( +[rank11]: ^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank11]: losses_reduced = forward_backward_func( +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: output_tensor, num_tokens = forward_step( +[rank0]: ^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank0]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank0]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank0]: batch = next(global_batches) +[rank0]: ^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank11]: output_tensor, num_tokens = forward_step( +[rank11]: ^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank11]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank11]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank0]: attention_mask = torch.ones( +[rank0]: ^^^^^^^^^^^ +[rank0]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 576.00 GiB. GPU 0 has a total capacity of 139.81 GiB of which 135.72 GiB is free. Including non-PyTorch memory, this process has 4.09 GiB memory in use. Of the allocated memory 2.52 GiB is allocated by PyTorch, and 91.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank11]: batch = next(global_batches) +[rank11]: ^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank11]: attention_mask = torch.ones( +[rank11]: ^^^^^^^^^^^ +[rank11]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 576.00 GiB. GPU 3 has a total capacity of 139.81 GiB of which 135.72 GiB is free. Including non-PyTorch memory, this process has 4.09 GiB memory in use. Of the allocated memory 2.52 GiB is allocated by PyTorch, and 91.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank6]: Traceback (most recent call last): +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank6]: pretrain( +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank6]: iteration, num_floating_point_operations_so_far = train( +[rank6]: ^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank6]: ) = train_step( +[rank6]: ^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank6]: losses_reduced = forward_backward_func( +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank12]: Traceback (most recent call last): +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank12]: pretrain( +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank12]: iteration, num_floating_point_operations_so_far = train( +[rank12]: ^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank12]: ) = train_step( +[rank12]: ^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank12]: losses_reduced = forward_backward_func( +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: output_tensor, num_tokens = forward_step( +[rank6]: ^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank6]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank6]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank6]: batch = next(global_batches) +[rank6]: ^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank12]: output_tensor, num_tokens = forward_step( +[rank12]: ^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank12]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank12]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank6]: attention_mask = torch.ones( +[rank6]: ^^^^^^^^^^^ +[rank6]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 576.00 GiB. GPU 6 has a total capacity of 139.81 GiB of which 135.72 GiB is free. Including non-PyTorch memory, this process has 4.09 GiB memory in use. Of the allocated memory 2.52 GiB is allocated by PyTorch, and 91.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank12]: batch = next(global_batches) +[rank12]: ^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank12]: attention_mask = torch.ones( +[rank12]: ^^^^^^^^^^^ +[rank12]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 576.00 GiB. GPU 4 has a total capacity of 139.81 GiB of which 135.70 GiB is free. Including non-PyTorch memory, this process has 4.10 GiB memory in use. Of the allocated memory 2.52 GiB is allocated by PyTorch, and 91.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank14]: Traceback (most recent call last): +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank14]: pretrain( +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank14]: iteration, num_floating_point_operations_so_far = train( +[rank14]: ^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank14]: ) = train_step( +[rank14]: ^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank14]: losses_reduced = forward_backward_func( +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank14]: output_tensor, num_tokens = forward_step( +[rank14]: ^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank14]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank14]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank14]: batch = next(global_batches) +[rank14]: ^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank14]: attention_mask = torch.ones( +[rank14]: ^^^^^^^^^^^ +[rank14]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 576.00 GiB. GPU 6 has a total capacity of 139.81 GiB of which 135.70 GiB is free. Including non-PyTorch memory, this process has 4.10 GiB memory in use. Of the allocated memory 2.52 GiB is allocated by PyTorch, and 91.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank10]: Traceback (most recent call last): +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank10]: pretrain( +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank10]: iteration, num_floating_point_operations_so_far = train( +[rank10]: ^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank10]: ) = train_step( +[rank10]: ^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank10]: losses_reduced = forward_backward_func( +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank10]: output_tensor, num_tokens = forward_step( +[rank10]: ^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank10]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank10]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank10]: batch = next(global_batches) +[rank10]: ^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank10]: attention_mask = torch.ones( +[rank10]: ^^^^^^^^^^^ +[rank10]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 576.00 GiB. GPU 2 has a total capacity of 139.81 GiB of which 135.70 GiB is free. Including non-PyTorch memory, this process has 4.10 GiB memory in use. Of the allocated memory 2.52 GiB is allocated by PyTorch, and 91.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank1]:[W621 22:16:05.303589027 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank5]:[W621 22:16:05.376770414 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank15]:[W621 22:16:05.505818746 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank9]:[W621 22:16:05.546801512 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank13]:[W621 22:16:05.587087594 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank11]:[W621 22:16:05.597153781 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank7]:[W621 22:16:05.528465532 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank3]:[W621 22:16:05.558725877 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +W0621 22:16:06.872000 1080631 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1080720 closing signal SIGTERM +W0621 22:16:06.874000 1080631 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1080722 closing signal SIGTERM +W0621 22:16:06.878000 1080631 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1080723 closing signal SIGTERM +W0621 22:16:06.879000 1080631 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1080724 closing signal SIGTERM +W0621 22:16:06.882000 1080631 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1080726 closing signal SIGTERM +W0621 22:16:06.886000 1080631 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1080727 closing signal SIGTERM +W0621 22:16:06.991000 2003918 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2003989 closing signal SIGTERM +W0621 22:16:06.994000 2003918 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2003991 closing signal SIGTERM +W0621 22:16:06.997000 2003918 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2003992 closing signal SIGTERM +W0621 22:16:06.997000 2003918 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2003993 closing signal SIGTERM +W0621 22:16:07.000000 2003918 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2003995 closing signal SIGTERM +E0621 22:16:08.749000 2003918 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 1 (pid: 2003990) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +Traceback (most recent call last): + File "", line 198, in _run_module_as_main + File "", line 88, in _run_code + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in + main() + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper + return arg(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main + launch(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch + run(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run + elastic_launch( + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__ + return launch_agent(self._config, self._entrypoint, list(args)) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent + raise ChildFailedError( +torch.distributed.elastic.multiprocessing.errors.ChildFailedError: +============================================================ +./pretrain_gpt_profile.py FAILED +------------------------------------------------------------ +Failures: +[1]: + time : 2025-06-21_22:16:06 + host : fs-mbz-gpu-286 + rank : 13 (local_rank: 5) + exitcode : 1 (pid: 2003994) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +------------------------------------------------------------ +Root Cause (first observed failure): +[0]: + time : 2025-06-21_22:16:06 + host : fs-mbz-gpu-286 + rank : 9 (local_rank: 1) + exitcode : 1 (pid: 2003990) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +============================================================ ++ set +x +E0621 22:16:09.423000 1080631 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 1 (pid: 1080721) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +Traceback (most recent call last): + File "", line 198, in _run_module_as_main + File "", line 88, in _run_code + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in + main() + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper + return arg(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main + launch(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch + run(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run + elastic_launch( + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__ + return launch_agent(self._config, self._entrypoint, list(args)) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent + raise ChildFailedError( +torch.distributed.elastic.multiprocessing.errors.ChildFailedError: +============================================================ +./pretrain_gpt_profile.py FAILED +------------------------------------------------------------ +Failures: +[1]: + time : 2025-06-21_22:16:06 + host : fs-mbz-gpu-239 + rank : 5 (local_rank: 5) + exitcode : 1 (pid: 1080725) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +------------------------------------------------------------ +Root Cause (first observed failure): +[0]: + time : 2025-06-21_22:16:06 + host : fs-mbz-gpu-239 + rank : 1 (local_rank: 1) + exitcode : 1 (pid: 1080721) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +============================================================ ++ set +x ++ for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072 ++ export PROF_CTX_LENGTH=16384 ++ PROF_CTX_LENGTH=16384 ++ name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L16384*tp2.cp8.bs16.json' ++ '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L16384*tp2.cp8.bs16.json' ']' ++ echo 'Running ctx_length=16384, TP_SIZE=2, CP_SIZE=8, BATCH_SIZE=16' ++ srun bash ./attnserver.sh ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 0 --rdzv_id 343241 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-239:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 16384 --max-position-embeddings 16384 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 1 --rdzv_id 343241 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-239:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 16384 --max-position-embeddings 16384 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:16:12.434000 2005763 site-packages/torch/distributed/run.py:766] +W0621 22:16:12.434000 2005763 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:16:12.434000 2005763 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:16:12.434000 2005763 site-packages/torch/distributed/run.py:766] ***************************************** +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:16:12.522000 1082546 site-packages/torch/distributed/run.py:766] +W0621 22:16:12.522000 1082546 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:16:12.522000 1082546 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:16:12.522000 1082546 site-packages/torch/distributed/run.py:766] ***************************************** +[rank5]:[W621 22:16:35.414282233 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 5] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank3]:[W621 22:16:35.414290557 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 3] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank1]:[W621 22:16:35.414310846 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 1] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank7]:[W621 22:16:35.414383635 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 7] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank13]:[W621 22:16:35.545869609 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 13] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank11]:[W621 22:16:35.545991896 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 11] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank15]:[W621 22:16:35.546165762 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 15] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank6]:[W621 22:16:35.423416164 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 6] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank4]:[W621 22:16:35.423454830 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 4] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank2]:[W621 22:16:35.424134402 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 2] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank14]:[W621 22:16:35.561011019 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 14] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank12]:[W621 22:16:35.561116619 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 12] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank10]:[W621 22:16:35.561116871 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 10] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank9]:[W621 22:16:35.561487160 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 9] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank8]:[W621 22:16:35.645401468 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 8] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank0]:[W621 22:16:35.579742214 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 0] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +[rank5]: Traceback (most recent call last): +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank5]: pretrain( +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank5]: iteration, num_floating_point_operations_so_far = train( +[rank5]: ^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank5]: ) = train_step( +[rank5]: ^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank5]: losses_reduced = forward_backward_func( +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank5]: output_tensor, num_tokens = forward_step( +[rank5]: ^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank5]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank5]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank5]: batch = next(global_batches) +[rank5]: ^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank5]: attention_mask = torch.ones( +[rank5]: ^^^^^^^^^^^ +[rank5]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1024.00 GiB. GPU 5 has a total capacity of 139.81 GiB of which 135.57 GiB is free. Including non-PyTorch memory, this process has 4.24 GiB memory in use. Of the allocated memory 2.66 GiB is allocated by PyTorch, and 81.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank7]: Traceback (most recent call last): +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank7]: pretrain( +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank7]: iteration, num_floating_point_operations_so_far = train( +[rank7]: ^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank7]: ) = train_step( +[rank7]: ^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank7]: losses_reduced = forward_backward_func( +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank7]: output_tensor, num_tokens = forward_step( +[rank7]: ^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank7]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank7]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank7]: batch = next(global_batches) +[rank7]: ^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank7]: attention_mask = torch.ones( +[rank7]: ^^^^^^^^^^^ +[rank7]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1024.00 GiB. GPU 7 has a total capacity of 139.81 GiB of which 135.57 GiB is free. Including non-PyTorch memory, this process has 4.24 GiB memory in use. Of the allocated memory 2.66 GiB is allocated by PyTorch, and 81.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank9]: Traceback (most recent call last): +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank9]: pretrain( +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank9]: iteration, num_floating_point_operations_so_far = train( +[rank9]: ^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank9]: ) = train_step( +[rank9]: ^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank9]: losses_reduced = forward_backward_func( +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank9]: output_tensor, num_tokens = forward_step( +[rank9]: ^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank9]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank9]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank9]: batch = next(global_batches) +[rank9]: ^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank9]: attention_mask = torch.ones( +[rank9]: ^^^^^^^^^^^ +[rank9]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1024.00 GiB. GPU 1 has a total capacity of 139.81 GiB of which 135.58 GiB is free. Including non-PyTorch memory, this process has 4.22 GiB memory in use. Of the allocated memory 2.66 GiB is allocated by PyTorch, and 81.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank2]: Traceback (most recent call last): +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank2]: pretrain( +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank2]: iteration, num_floating_point_operations_so_far = train( +[rank2]: ^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank2]: ) = train_step( +[rank2]: ^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank2]: losses_reduced = forward_backward_func( +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank2]: output_tensor, num_tokens = forward_step( +[rank2]: ^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank2]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank2]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank2]: batch = next(global_batches) +[rank2]: ^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank2]: attention_mask = torch.ones( +[rank2]: ^^^^^^^^^^^ +[rank2]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1024.00 GiB. GPU 2 has a total capacity of 139.81 GiB of which 135.58 GiB is free. Including non-PyTorch memory, this process has 4.22 GiB memory in use. Of the allocated memory 2.66 GiB is allocated by PyTorch, and 81.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank13]: Traceback (most recent call last): +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank13]: pretrain( +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank13]: iteration, num_floating_point_operations_so_far = train( +[rank13]: ^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank13]: ) = train_step( +[rank13]: ^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank13]: losses_reduced = forward_backward_func( +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank13]: output_tensor, num_tokens = forward_step( +[rank13]: ^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank13]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank13]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank13]: batch = next(global_batches) +[rank13]: ^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank13]: attention_mask = torch.ones( +[rank13]: ^^^^^^^^^^^ +[rank13]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1024.00 GiB. GPU 5 has a total capacity of 139.81 GiB of which 135.58 GiB is free. Including non-PyTorch memory, this process has 4.22 GiB memory in use. Of the allocated memory 2.66 GiB is allocated by PyTorch, and 81.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank1]: Traceback (most recent call last): +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank1]: pretrain( +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank1]: iteration, num_floating_point_operations_so_far = train( +[rank1]: ^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank1]: ) = train_step( +[rank1]: ^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank1]: losses_reduced = forward_backward_func( +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank1]: output_tensor, num_tokens = forward_step( +[rank1]: ^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank1]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank1]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank1]: batch = next(global_batches) +[rank1]: ^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank1]: attention_mask = torch.ones( +[rank1]: ^^^^^^^^^^^ +[rank1]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1024.00 GiB. GPU 1 has a total capacity of 139.81 GiB of which 135.57 GiB is free. Including non-PyTorch memory, this process has 4.24 GiB memory in use. Of the allocated memory 2.66 GiB is allocated by PyTorch, and 81.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank4]: Traceback (most recent call last): +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank4]: pretrain( +[rank8]: Traceback (most recent call last): +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank8]: pretrain( +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank8]: iteration, num_floating_point_operations_so_far = train( +[rank8]: ^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank8]: ) = train_step( +[rank8]: ^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank8]: losses_reduced = forward_backward_func( +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank4]: iteration, num_floating_point_operations_so_far = train( +[rank4]: ^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank4]: ) = train_step( +[rank4]: ^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank4]: losses_reduced = forward_backward_func( +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank4]: output_tensor, num_tokens = forward_step( +[rank4]: ^^^^^^^^^^^^^ +[rank8]: output_tensor, num_tokens = forward_step( +[rank8]: ^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank8]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank8]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank8]: batch = next(global_batches) +[rank8]: ^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank8]: attention_mask = torch.ones( +[rank8]: ^^^^^^^^^^^ +[rank8]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1024.00 GiB. GPU 0 has a total capacity of 139.81 GiB of which 135.57 GiB is free. Including non-PyTorch memory, this process has 4.24 GiB memory in use. Of the allocated memory 2.66 GiB is allocated by PyTorch, and 81.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank4]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank4]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank4]: batch = next(global_batches) +[rank4]: ^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank4]: attention_mask = torch.ones( +[rank4]: ^^^^^^^^^^^ +[rank11]: Traceback (most recent call last): +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank11]: pretrain( +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank11]: iteration, num_floating_point_operations_so_far = train( +[rank11]: ^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank11]: ) = train_step( +[rank11]: ^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank11]: losses_reduced = forward_backward_func( +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1024.00 GiB. GPU 4 has a total capacity of 139.81 GiB of which 135.58 GiB is free. Including non-PyTorch memory, this process has 4.22 GiB memory in use. Of the allocated memory 2.66 GiB is allocated by PyTorch, and 81.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank11]: output_tensor, num_tokens = forward_step( +[rank11]: ^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank11]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank11]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: Traceback (most recent call last): +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank3]: pretrain( +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank3]: iteration, num_floating_point_operations_so_far = train( +[rank3]: ^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank3]: ) = train_step( +[rank3]: ^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank3]: losses_reduced = forward_backward_func( +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank11]: batch = next(global_batches) +[rank11]: ^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank11]: attention_mask = torch.ones( +[rank11]: ^^^^^^^^^^^ +[rank11]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1024.00 GiB. GPU 3 has a total capacity of 139.81 GiB of which 135.58 GiB is free. Including non-PyTorch memory, this process has 4.22 GiB memory in use. Of the allocated memory 2.66 GiB is allocated by PyTorch, and 81.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank3]: output_tensor, num_tokens = forward_step( +[rank3]: ^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank3]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank3]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank3]: batch = next(global_batches) +[rank3]: ^^^^^^^^^^^^^^^^^^^^ +[rank15]: Traceback (most recent call last): +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank15]: pretrain( +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank15]: iteration, num_floating_point_operations_so_far = train( +[rank15]: ^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank15]: ) = train_step( +[rank15]: ^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank15]: losses_reduced = forward_backward_func( +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank3]: attention_mask = torch.ones( +[rank3]: ^^^^^^^^^^^ +[rank3]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1024.00 GiB. GPU 3 has a total capacity of 139.81 GiB of which 135.57 GiB is free. Including non-PyTorch memory, this process has 4.24 GiB memory in use. Of the allocated memory 2.66 GiB is allocated by PyTorch, and 81.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank15]: output_tensor, num_tokens = forward_step( +[rank15]: ^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank15]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank15]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: Traceback (most recent call last): +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank0]: pretrain( +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank0]: iteration, num_floating_point_operations_so_far = train( +[rank0]: ^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank0]: ) = train_step( +[rank0]: ^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank0]: losses_reduced = forward_backward_func( +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank15]: batch = next(global_batches) +[rank15]: ^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank15]: attention_mask = torch.ones( +[rank15]: ^^^^^^^^^^^ +[rank15]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1024.00 GiB. GPU 7 has a total capacity of 139.81 GiB of which 135.58 GiB is free. Including non-PyTorch memory, this process has 4.22 GiB memory in use. Of the allocated memory 2.66 GiB is allocated by PyTorch, and 81.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank0]: output_tensor, num_tokens = forward_step( +[rank0]: ^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank0]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank0]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank0]: batch = next(global_batches) +[rank0]: ^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank0]: attention_mask = torch.ones( +[rank0]: ^^^^^^^^^^^ +[rank0]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1024.00 GiB. GPU 0 has a total capacity of 139.81 GiB of which 135.58 GiB is free. Including non-PyTorch memory, this process has 4.22 GiB memory in use. Of the allocated memory 2.66 GiB is allocated by PyTorch, and 81.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank6]: Traceback (most recent call last): +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank6]: pretrain( +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank6]: iteration, num_floating_point_operations_so_far = train( +[rank6]: ^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank6]: ) = train_step( +[rank6]: ^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank6]: losses_reduced = forward_backward_func( +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank6]: output_tensor, num_tokens = forward_step( +[rank6]: ^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank6]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank6]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank6]: batch = next(global_batches) +[rank6]: ^^^^^^^^^^^^^^^^^^^^ +[rank12]: Traceback (most recent call last): +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank12]: pretrain( +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank12]: iteration, num_floating_point_operations_so_far = train( +[rank12]: ^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank12]: ) = train_step( +[rank12]: ^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank12]: losses_reduced = forward_backward_func( +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank6]: attention_mask = torch.ones( +[rank6]: ^^^^^^^^^^^ +[rank6]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1024.00 GiB. GPU 6 has a total capacity of 139.81 GiB of which 135.58 GiB is free. Including non-PyTorch memory, this process has 4.22 GiB memory in use. Of the allocated memory 2.66 GiB is allocated by PyTorch, and 81.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank12]: output_tensor, num_tokens = forward_step( +[rank12]: ^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank12]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank12]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank12]: batch = next(global_batches) +[rank12]: ^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank12]: attention_mask = torch.ones( +[rank12]: ^^^^^^^^^^^ +[rank12]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1024.00 GiB. GPU 4 has a total capacity of 139.81 GiB of which 135.57 GiB is free. Including non-PyTorch memory, this process has 4.24 GiB memory in use. Of the allocated memory 2.66 GiB is allocated by PyTorch, and 81.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank10]: Traceback (most recent call last): +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank10]: pretrain( +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank10]: iteration, num_floating_point_operations_so_far = train( +[rank10]: ^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank10]: ) = train_step( +[rank10]: ^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank10]: losses_reduced = forward_backward_func( +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank10]: output_tensor, num_tokens = forward_step( +[rank10]: ^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank10]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank10]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank10]: batch = next(global_batches) +[rank10]: ^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank10]: attention_mask = torch.ones( +[rank10]: ^^^^^^^^^^^ +[rank10]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1024.00 GiB. GPU 2 has a total capacity of 139.81 GiB of which 135.57 GiB is free. Including non-PyTorch memory, this process has 4.24 GiB memory in use. Of the allocated memory 2.66 GiB is allocated by PyTorch, and 81.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank14]: Traceback (most recent call last): +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank14]: pretrain( +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank14]: iteration, num_floating_point_operations_so_far = train( +[rank14]: ^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank14]: ) = train_step( +[rank14]: ^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank14]: losses_reduced = forward_backward_func( +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank14]: output_tensor, num_tokens = forward_step( +[rank14]: ^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank14]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank14]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank14]: batch = next(global_batches) +[rank14]: ^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank14]: attention_mask = torch.ones( +[rank14]: ^^^^^^^^^^^ +[rank14]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1024.00 GiB. GPU 6 has a total capacity of 139.81 GiB of which 135.57 GiB is free. Including non-PyTorch memory, this process has 4.24 GiB memory in use. Of the allocated memory 2.66 GiB is allocated by PyTorch, and 81.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank13]:[W621 22:16:45.322681033 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank5]:[W621 22:16:45.254404046 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank1]:[W621 22:16:45.341632703 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank3]:[W621 22:16:45.365135844 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank7]:[W621 22:16:45.365224849 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank15]:[W621 22:16:45.615827161 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank9]:[W621 22:16:45.807040180 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank11]:[W621 22:16:45.847486068 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +W0621 22:16:46.655000 2005763 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2005835 closing signal SIGTERM +W0621 22:16:46.657000 2005763 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2005836 closing signal SIGTERM +W0621 22:16:46.658000 2005763 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2005837 closing signal SIGTERM +W0621 22:16:46.661000 2005763 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2005838 closing signal SIGTERM +W0621 22:16:46.662000 2005763 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2005839 closing signal SIGTERM +W0621 22:16:46.665000 2005763 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2005841 closing signal SIGTERM +W0621 22:16:46.667000 2005763 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2005842 closing signal SIGTERM +W0621 22:16:46.749000 1082546 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1082617 closing signal SIGTERM +W0621 22:16:46.752000 1082546 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1082618 closing signal SIGTERM +W0621 22:16:46.753000 1082546 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1082619 closing signal SIGTERM +W0621 22:16:46.755000 1082546 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1082620 closing signal SIGTERM +W0621 22:16:46.755000 1082546 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1082621 closing signal SIGTERM +W0621 22:16:46.759000 1082546 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1082623 closing signal SIGTERM +W0621 22:16:46.777000 1082546 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1082624 closing signal SIGTERM +E0621 22:16:48.073000 2005763 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 5 (pid: 2005840) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +Traceback (most recent call last): + File "", line 198, in _run_module_as_main + File "", line 88, in _run_code + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in + main() + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper + return arg(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main + launch(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch + run(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run + elastic_launch( + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__ + return launch_agent(self._config, self._entrypoint, list(args)) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent + raise ChildFailedError( +torch.distributed.elastic.multiprocessing.errors.ChildFailedError: +============================================================ +./pretrain_gpt_profile.py FAILED +------------------------------------------------------------ +Failures: + +------------------------------------------------------------ +Root Cause (first observed failure): +[0]: + time : 2025-06-21_22:16:46 + host : fs-mbz-gpu-286 + rank : 13 (local_rank: 5) + exitcode : 1 (pid: 2005840) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +============================================================ ++ set +x +E0621 22:16:48.461000 1082546 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 5 (pid: 1082622) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +Traceback (most recent call last): + File "", line 198, in _run_module_as_main + File "", line 88, in _run_code + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in + main() + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper + return arg(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main + launch(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch + run(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run + elastic_launch( + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__ + return launch_agent(self._config, self._entrypoint, list(args)) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent + raise ChildFailedError( +torch.distributed.elastic.multiprocessing.errors.ChildFailedError: +============================================================ +./pretrain_gpt_profile.py FAILED +------------------------------------------------------------ +Failures: + +------------------------------------------------------------ +Root Cause (first observed failure): +[0]: + time : 2025-06-21_22:16:46 + host : fs-mbz-gpu-239 + rank : 5 (local_rank: 5) + exitcode : 1 (pid: 1082622) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +============================================================ ++ set +x ++ for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072 ++ export PROF_CTX_LENGTH=24576 ++ PROF_CTX_LENGTH=24576 ++ name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L24576*tp2.cp8.bs16.json' ++ '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L24576*tp2.cp8.bs16.json' ']' ++ echo 'Running ctx_length=24576, TP_SIZE=2, CP_SIZE=8, BATCH_SIZE=16' ++ srun bash ./attnserver.sh ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 0 --rdzv_id 343241 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-239:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 24576 --max-position-embeddings 24576 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 1 --rdzv_id 343241 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-239:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 24576 --max-position-embeddings 24576 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:16:51.491000 1084460 site-packages/torch/distributed/run.py:766] +W0621 22:16:51.491000 1084460 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:16:51.491000 1084460 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:16:51.491000 1084460 site-packages/torch/distributed/run.py:766] ***************************************** +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:16:51.510000 2007625 site-packages/torch/distributed/run.py:766] +W0621 22:16:51.510000 2007625 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:16:51.510000 2007625 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:16:51.510000 2007625 site-packages/torch/distributed/run.py:766] ***************************************** +[rank4]:[W621 22:17:13.777504443 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 4] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank6]:[W621 22:17:13.777503544 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 6] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank3]:[W621 22:17:13.781086920 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 3] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank2]:[W621 22:17:13.781240358 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 2] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank5]:[W621 22:17:13.781866049 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 5] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank7]:[W621 22:17:13.782604901 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 7] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank10]:[W621 22:17:13.913924965 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 10] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank11]:[W621 22:17:13.913927274 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 11] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank14]:[W621 22:17:13.913954421 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 14] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank9]:[W621 22:17:13.913988998 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 9] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank1]:[W621 22:17:13.784768088 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 1] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank13]:[W621 22:17:13.914032045 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 13] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank15]:[W621 22:17:13.914047579 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 15] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank12]:[W621 22:17:13.915153862 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 12] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank8]:[W621 22:17:13.997420817 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 8] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank0]:[W621 22:17:13.924839053 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 0] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +[rank14]: Traceback (most recent call last): +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank14]: pretrain( +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank14]: iteration, num_floating_point_operations_so_far = train( +[rank14]: ^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank14]: ) = train_step( +[rank14]: ^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank14]: losses_reduced = forward_backward_func( +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank14]: output_tensor, num_tokens = forward_step( +[rank14]: ^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank14]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank14]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank14]: batch = next(global_batches) +[rank14]: ^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank14]: attention_mask = torch.ones( +[rank14]: ^^^^^^^^^^^ +[rank14]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 2304.00 GiB. GPU 6 has a total capacity of 139.81 GiB of which 135.27 GiB is free. Including non-PyTorch memory, this process has 4.53 GiB memory in use. Of the allocated memory 2.95 GiB is allocated by PyTorch, and 89.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank10]: Traceback (most recent call last): +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank10]: pretrain( +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank10]: iteration, num_floating_point_operations_so_far = train( +[rank10]: ^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank10]: ) = train_step( +[rank10]: ^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank10]: losses_reduced = forward_backward_func( +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank10]: output_tensor, num_tokens = forward_step( +[rank10]: ^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank10]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank10]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: Traceback (most recent call last): +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank3]: pretrain( +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank3]: iteration, num_floating_point_operations_so_far = train( +[rank3]: ^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank3]: ) = train_step( +[rank3]: ^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank3]: losses_reduced = forward_backward_func( +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank10]: batch = next(global_batches) +[rank10]: ^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank10]: attention_mask = torch.ones( +[rank10]: ^^^^^^^^^^^ +[rank10]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 2304.00 GiB. GPU 2 has a total capacity of 139.81 GiB of which 135.27 GiB is free. Including non-PyTorch memory, this process has 4.53 GiB memory in use. Of the allocated memory 2.95 GiB is allocated by PyTorch, and 89.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank3]: output_tensor, num_tokens = forward_step( +[rank3]: ^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank3]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank3]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank3]: batch = next(global_batches) +[rank3]: ^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank3]: attention_mask = torch.ones( +[rank3]: ^^^^^^^^^^^ +[rank3]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 2304.00 GiB. GPU 3 has a total capacity of 139.81 GiB of which 135.27 GiB is free. Including non-PyTorch memory, this process has 4.53 GiB memory in use. Of the allocated memory 2.95 GiB is allocated by PyTorch, and 89.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank4]: Traceback (most recent call last): +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank4]: pretrain( +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank4]: iteration, num_floating_point_operations_so_far = train( +[rank4]: ^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank4]: ) = train_step( +[rank4]: ^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank4]: losses_reduced = forward_backward_func( +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank4]: output_tensor, num_tokens = forward_step( +[rank4]: ^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank4]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank4]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank4]: batch = next(global_batches) +[rank4]: ^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank4]: attention_mask = torch.ones( +[rank4]: ^^^^^^^^^^^ +[rank4]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 2304.00 GiB. GPU 4 has a total capacity of 139.81 GiB of which 135.29 GiB is free. Including non-PyTorch memory, this process has 4.52 GiB memory in use. Of the allocated memory 2.95 GiB is allocated by PyTorch, and 89.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank9]: Traceback (most recent call last): +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank9]: pretrain( +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank9]: iteration, num_floating_point_operations_so_far = train( +[rank9]: ^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank9]: ) = train_step( +[rank9]: ^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank9]: losses_reduced = forward_backward_func( +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank9]: output_tensor, num_tokens = forward_step( +[rank9]: ^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank9]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank9]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank9]: batch = next(global_batches) +[rank9]: ^^^^^^^^^^^^^^^^^^^^ +[rank2]: Traceback (most recent call last): +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank2]: pretrain( +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank2]: iteration, num_floating_point_operations_so_far = train( +[rank2]: ^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank2]: ) = train_step( +[rank2]: ^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank2]: losses_reduced = forward_backward_func( +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank9]: attention_mask = torch.ones( +[rank9]: ^^^^^^^^^^^ +[rank9]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 2304.00 GiB. GPU 1 has a total capacity of 139.81 GiB of which 135.29 GiB is free. Including non-PyTorch memory, this process has 4.52 GiB memory in use. Of the allocated memory 2.95 GiB is allocated by PyTorch, and 89.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank11]: Traceback (most recent call last): +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank11]: pretrain( +[rank2]: output_tensor, num_tokens = forward_step( +[rank2]: ^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank2]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank2]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank2]: batch = next(global_batches) +[rank2]: ^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank11]: iteration, num_floating_point_operations_so_far = train( +[rank11]: ^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank11]: ) = train_step( +[rank11]: ^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank11]: losses_reduced = forward_backward_func( +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank11]: output_tensor, num_tokens = forward_step( +[rank11]: ^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank2]: attention_mask = torch.ones( +[rank2]: ^^^^^^^^^^^ +[rank2]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 2304.00 GiB. GPU 2 has a total capacity of 139.81 GiB of which 135.29 GiB is free. Including non-PyTorch memory, this process has 4.52 GiB memory in use. Of the allocated memory 2.95 GiB is allocated by PyTorch, and 89.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank11]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank11]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank11]: batch = next(global_batches) +[rank11]: ^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank11]: attention_mask = torch.ones( +[rank11]: ^^^^^^^^^^^ +[rank11]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 2304.00 GiB. GPU 3 has a total capacity of 139.81 GiB of which 135.29 GiB is free. Including non-PyTorch memory, this process has 4.52 GiB memory in use. Of the allocated memory 2.95 GiB is allocated by PyTorch, and 89.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank13]: Traceback (most recent call last): +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank13]: pretrain( +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank13]: iteration, num_floating_point_operations_so_far = train( +[rank13]: ^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank13]: ) = train_step( +[rank13]: ^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank13]: losses_reduced = forward_backward_func( +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank13]: output_tensor, num_tokens = forward_step( +[rank13]: ^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank13]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank13]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank13]: batch = next(global_batches) +[rank13]: ^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank13]: attention_mask = torch.ones( +[rank13]: ^^^^^^^^^^^ +[rank13]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 2304.00 GiB. GPU 5 has a total capacity of 139.81 GiB of which 135.29 GiB is free. Including non-PyTorch memory, this process has 4.52 GiB memory in use. Of the allocated memory 2.95 GiB is allocated by PyTorch, and 89.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank12]: Traceback (most recent call last): +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank12]: pretrain( +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank12]: iteration, num_floating_point_operations_so_far = train( +[rank12]: ^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank12]: ) = train_step( +[rank12]: ^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank12]: losses_reduced = forward_backward_func( +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank12]: output_tensor, num_tokens = forward_step( +[rank12]: ^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank12]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank12]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: Traceback (most recent call last): +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank5]: pretrain( +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank5]: iteration, num_floating_point_operations_so_far = train( +[rank5]: ^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank5]: ) = train_step( +[rank5]: ^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank5]: losses_reduced = forward_backward_func( +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank12]: batch = next(global_batches) +[rank12]: ^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank12]: attention_mask = torch.ones( +[rank12]: ^^^^^^^^^^^ +[rank12]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 2304.00 GiB. GPU 4 has a total capacity of 139.81 GiB of which 135.27 GiB is free. Including non-PyTorch memory, this process has 4.53 GiB memory in use. Of the allocated memory 2.95 GiB is allocated by PyTorch, and 89.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank5]: output_tensor, num_tokens = forward_step( +[rank5]: ^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank5]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank5]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank5]: batch = next(global_batches) +[rank5]: ^^^^^^^^^^^^^^^^^^^^ +[rank15]: Traceback (most recent call last): +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank15]: pretrain( +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank15]: iteration, num_floating_point_operations_so_far = train( +[rank15]: ^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank15]: ) = train_step( +[rank15]: ^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank15]: losses_reduced = forward_backward_func( +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank5]: attention_mask = torch.ones( +[rank5]: ^^^^^^^^^^^ +[rank5]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 2304.00 GiB. GPU 5 has a total capacity of 139.81 GiB of which 135.27 GiB is free. Including non-PyTorch memory, this process has 4.53 GiB memory in use. Of the allocated memory 2.95 GiB is allocated by PyTorch, and 89.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank15]: output_tensor, num_tokens = forward_step( +[rank15]: ^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank15]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank15]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank15]: batch = next(global_batches) +[rank15]: ^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank15]: attention_mask = torch.ones( +[rank15]: ^^^^^^^^^^^ +[rank15]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 2304.00 GiB. GPU 7 has a total capacity of 139.81 GiB of which 135.29 GiB is free. Including non-PyTorch memory, this process has 4.52 GiB memory in use. Of the allocated memory 2.95 GiB is allocated by PyTorch, and 89.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank8]: Traceback (most recent call last): +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank8]: pretrain( +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank8]: iteration, num_floating_point_operations_so_far = train( +[rank8]: ^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank8]: ) = train_step( +[rank8]: ^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank8]: losses_reduced = forward_backward_func( +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank8]: output_tensor, num_tokens = forward_step( +[rank8]: ^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank8]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank8]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank8]: batch = next(global_batches) +[rank8]: ^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank8]: attention_mask = torch.ones( +[rank8]: ^^^^^^^^^^^ +[rank8]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 2304.00 GiB. GPU 0 has a total capacity of 139.81 GiB of which 135.27 GiB is free. Including non-PyTorch memory, this process has 4.53 GiB memory in use. Of the allocated memory 2.95 GiB is allocated by PyTorch, and 89.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank1]: Traceback (most recent call last): +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank1]: pretrain( +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank1]: iteration, num_floating_point_operations_so_far = train( +[rank1]: ^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank1]: ) = train_step( +[rank1]: ^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank1]: losses_reduced = forward_backward_func( +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank1]: output_tensor, num_tokens = forward_step( +[rank1]: ^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank1]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank1]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank1]: batch = next(global_batches) +[rank1]: ^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank1]: attention_mask = torch.ones( +[rank1]: ^^^^^^^^^^^ +[rank1]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 2304.00 GiB. GPU 1 has a total capacity of 139.81 GiB of which 135.27 GiB is free. Including non-PyTorch memory, this process has 4.53 GiB memory in use. Of the allocated memory 2.95 GiB is allocated by PyTorch, and 89.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank0]: Traceback (most recent call last): +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank0]: pretrain( +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank0]: iteration, num_floating_point_operations_so_far = train( +[rank0]: ^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank0]: ) = train_step( +[rank0]: ^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank0]: losses_reduced = forward_backward_func( +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank0]: output_tensor, num_tokens = forward_step( +[rank0]: ^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank0]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank0]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank0]: batch = next(global_batches) +[rank0]: ^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank0]: attention_mask = torch.ones( +[rank0]: ^^^^^^^^^^^ +[rank0]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 2304.00 GiB. GPU 0 has a total capacity of 139.81 GiB of which 135.29 GiB is free. Including non-PyTorch memory, this process has 4.52 GiB memory in use. Of the allocated memory 2.95 GiB is allocated by PyTorch, and 89.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank7]: Traceback (most recent call last): +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank7]: pretrain( +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank7]: iteration, num_floating_point_operations_so_far = train( +[rank7]: ^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank7]: ) = train_step( +[rank7]: ^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank7]: losses_reduced = forward_backward_func( +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank7]: output_tensor, num_tokens = forward_step( +[rank7]: ^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank7]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank7]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank7]: batch = next(global_batches) +[rank7]: ^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank7]: attention_mask = torch.ones( +[rank7]: ^^^^^^^^^^^ +[rank7]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 2304.00 GiB. GPU 7 has a total capacity of 139.81 GiB of which 135.27 GiB is free. Including non-PyTorch memory, this process has 4.53 GiB memory in use. Of the allocated memory 2.95 GiB is allocated by PyTorch, and 89.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank6]: Traceback (most recent call last): +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank6]: pretrain( +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank6]: iteration, num_floating_point_operations_so_far = train( +[rank6]: ^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank6]: ) = train_step( +[rank6]: ^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank6]: losses_reduced = forward_backward_func( +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank6]: output_tensor, num_tokens = forward_step( +[rank6]: ^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank6]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank6]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank6]: batch = next(global_batches) +[rank6]: ^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank6]: attention_mask = torch.ones( +[rank6]: ^^^^^^^^^^^ +[rank6]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 2304.00 GiB. GPU 6 has a total capacity of 139.81 GiB of which 135.29 GiB is free. Including non-PyTorch memory, this process has 4.52 GiB memory in use. Of the allocated memory 2.95 GiB is allocated by PyTorch, and 89.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank5]:[W621 22:17:23.768291771 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank7]:[W621 22:17:23.778634682 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank3]:[W621 22:17:23.784744887 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank9]:[W621 22:17:23.913935633 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank13]:[W621 22:17:23.927767527 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank11]:[W621 22:17:23.932072922 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank15]:[W621 22:17:23.933996663 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank1]:[W621 22:17:24.068057448 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +W0621 22:17:25.395000 1084460 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1084532 closing signal SIGTERM +W0621 22:17:25.398000 1084460 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1084533 closing signal SIGTERM +W0621 22:17:25.398000 1084460 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1084534 closing signal SIGTERM +W0621 22:17:25.401000 1084460 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1084535 closing signal SIGTERM +W0621 22:17:25.402000 1084460 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1084536 closing signal SIGTERM +W0621 22:17:25.404000 1084460 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1084538 closing signal SIGTERM +W0621 22:17:25.407000 1084460 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1084539 closing signal SIGTERM +W0621 22:17:25.429000 2007625 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2007696 closing signal SIGTERM +W0621 22:17:25.432000 2007625 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2007698 closing signal SIGTERM +W0621 22:17:25.435000 2007625 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2007699 closing signal SIGTERM +W0621 22:17:25.435000 2007625 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2007700 closing signal SIGTERM +W0621 22:17:25.437000 2007625 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2007701 closing signal SIGTERM +W0621 22:17:25.438000 2007625 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2007702 closing signal SIGTERM +E0621 22:17:25.743000 2007625 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 1 (pid: 2007697) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +Traceback (most recent call last): + File "", line 198, in _run_module_as_main + File "", line 88, in _run_code + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in + main() + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper + return arg(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main + launch(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch + run(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run + elastic_launch( + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__ + return launch_agent(self._config, self._entrypoint, list(args)) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent + raise ChildFailedError( +torch.distributed.elastic.multiprocessing.errors.ChildFailedError: +============================================================ +./pretrain_gpt_profile.py FAILED +------------------------------------------------------------ +Failures: +[1]: + time : 2025-06-21_22:17:25 + host : fs-mbz-gpu-286 + rank : 15 (local_rank: 7) + exitcode : 1 (pid: 2007703) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +------------------------------------------------------------ +Root Cause (first observed failure): +[0]: + time : 2025-06-21_22:17:25 + host : fs-mbz-gpu-286 + rank : 9 (local_rank: 1) + exitcode : 1 (pid: 2007697) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +============================================================ +E0621 22:17:25.810000 1084460 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 5 (pid: 1084537) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +Traceback (most recent call last): + File "", line 198, in _run_module_as_main + File "", line 88, in _run_code + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in + main() + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper + return arg(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main + launch(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch + run(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run + elastic_launch( + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__ + return launch_agent(self._config, self._entrypoint, list(args)) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent + raise ChildFailedError( +torch.distributed.elastic.multiprocessing.errors.ChildFailedError: +============================================================ +./pretrain_gpt_profile.py FAILED +------------------------------------------------------------ +Failures: + +------------------------------------------------------------ +Root Cause (first observed failure): +[0]: + time : 2025-06-21_22:17:25 + host : fs-mbz-gpu-239 + rank : 5 (local_rank: 5) + exitcode : 1 (pid: 1084537) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +============================================================ ++ set +x ++ set +x ++ for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072 ++ export PROF_CTX_LENGTH=32768 ++ PROF_CTX_LENGTH=32768 ++ name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L32768*tp2.cp8.bs16.json' ++ '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L32768*tp2.cp8.bs16.json' ']' ++ echo 'Running ctx_length=32768, TP_SIZE=2, CP_SIZE=8, BATCH_SIZE=16' ++ srun bash ./attnserver.sh ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 1 --rdzv_id 343241 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-239:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 32768 --max-position-embeddings 32768 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 0 --rdzv_id 343241 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-239:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 32768 --max-position-embeddings 32768 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:17:29.879000 2009444 site-packages/torch/distributed/run.py:766] +W0621 22:17:29.879000 2009444 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:17:29.879000 2009444 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:17:29.879000 2009444 site-packages/torch/distributed/run.py:766] ***************************************** +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:17:29.988000 1086332 site-packages/torch/distributed/run.py:766] +W0621 22:17:29.988000 1086332 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:17:29.988000 1086332 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:17:29.988000 1086332 site-packages/torch/distributed/run.py:766] ***************************************** +[rank6]:[W621 22:17:53.565294653 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 6] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank4]:[W621 22:17:53.565299919 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 4] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank2]:[W621 22:17:53.565314980 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 2] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank14]:[W621 22:17:53.698525090 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 14] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank12]:[W621 22:17:53.698529888 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 12] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank10]:[W621 22:17:53.698770890 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 10] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank5]:[W621 22:17:53.578757073 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 5] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank7]:[W621 22:17:53.579584254 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 7] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank3]:[W621 22:17:53.582164280 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 3] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank1]:[W621 22:17:53.582162146 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 1] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank15]:[W621 22:17:53.711495856 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 15] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank9]:[W621 22:17:53.711566449 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 9] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank11]:[W621 22:17:53.711596895 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 11] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank13]:[W621 22:17:53.712700502 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 13] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank8]:[W621 22:17:53.795469327 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 8] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank0]:[W621 22:17:53.708769712 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 0] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +[rank11]: Traceback (most recent call last): +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank11]: pretrain( +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank11]: iteration, num_floating_point_operations_so_far = train( +[rank11]: ^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank11]: ) = train_step( +[rank11]: ^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank11]: losses_reduced = forward_backward_func( +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank11]: output_tensor, num_tokens = forward_step( +[rank11]: ^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank11]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank11]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank11]: batch = next(global_batches) +[rank11]: ^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank11]: attention_mask = torch.ones( +[rank11]: ^^^^^^^^^^^ +[rank11]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 4096.00 GiB. GPU 3 has a total capacity of 139.81 GiB of which 135.02 GiB is free. Including non-PyTorch memory, this process has 4.78 GiB memory in use. Of the allocated memory 3.24 GiB is allocated by PyTorch, and 65.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank13]: Traceback (most recent call last): +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank13]: pretrain( +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank13]: iteration, num_floating_point_operations_so_far = train( +[rank13]: ^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank13]: ) = train_step( +[rank13]: ^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank13]: losses_reduced = forward_backward_func( +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank13]: output_tensor, num_tokens = forward_step( +[rank13]: ^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank13]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank13]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank13]: batch = next(global_batches) +[rank13]: ^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank13]: attention_mask = torch.ones( +[rank13]: ^^^^^^^^^^^ +[rank13]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 4096.00 GiB. GPU 5 has a total capacity of 139.81 GiB of which 135.02 GiB is free. Including non-PyTorch memory, this process has 4.78 GiB memory in use. Of the allocated memory 3.24 GiB is allocated by PyTorch, and 65.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank8]: Traceback (most recent call last): +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank8]: pretrain( +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank8]: iteration, num_floating_point_operations_so_far = train( +[rank8]: ^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank8]: ) = train_step( +[rank8]: ^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank8]: losses_reduced = forward_backward_func( +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank8]: output_tensor, num_tokens = forward_step( +[rank8]: ^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank8]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank8]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank8]: batch = next(global_batches) +[rank8]: ^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank8]: attention_mask = torch.ones( +[rank8]: ^^^^^^^^^^^ +[rank8]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 4096.00 GiB. GPU 0 has a total capacity of 139.81 GiB of which 135.00 GiB is free. Including non-PyTorch memory, this process has 4.80 GiB memory in use. Of the allocated memory 3.24 GiB is allocated by PyTorch, and 65.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank15]: Traceback (most recent call last): +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank15]: pretrain( +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank15]: iteration, num_floating_point_operations_so_far = train( +[rank15]: ^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank15]: ) = train_step( +[rank15]: ^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank15]: losses_reduced = forward_backward_func( +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank15]: output_tensor, num_tokens = forward_step( +[rank15]: ^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank15]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank15]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank15]: batch = next(global_batches) +[rank15]: ^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank15]: attention_mask = torch.ones( +[rank15]: ^^^^^^^^^^^ +[rank15]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 4096.00 GiB. GPU 7 has a total capacity of 139.81 GiB of which 135.02 GiB is free. Including non-PyTorch memory, this process has 4.78 GiB memory in use. Of the allocated memory 3.24 GiB is allocated by PyTorch, and 65.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank3]: Traceback (most recent call last): +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank3]: pretrain( +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank3]: iteration, num_floating_point_operations_so_far = train( +[rank3]: ^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank3]: ) = train_step( +[rank3]: ^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank3]: losses_reduced = forward_backward_func( +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank3]: output_tensor, num_tokens = forward_step( +[rank3]: ^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank3]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank3]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank3]: batch = next(global_batches) +[rank3]: ^^^^^^^^^^^^^^^^^^^^ +[rank9]: Traceback (most recent call last): +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank9]: pretrain( +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank9]: iteration, num_floating_point_operations_so_far = train( +[rank9]: ^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank9]: ) = train_step( +[rank9]: ^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank9]: losses_reduced = forward_backward_func( +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank3]: attention_mask = torch.ones( +[rank3]: ^^^^^^^^^^^ +[rank3]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 4096.00 GiB. GPU 3 has a total capacity of 139.81 GiB of which 135.00 GiB is free. Including non-PyTorch memory, this process has 4.80 GiB memory in use. Of the allocated memory 3.24 GiB is allocated by PyTorch, and 65.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank9]: output_tensor, num_tokens = forward_step( +[rank9]: ^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank9]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank9]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank9]: batch = next(global_batches) +[rank9]: ^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank9]: attention_mask = torch.ones( +[rank9]: ^^^^^^^^^^^ +[rank9]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 4096.00 GiB. GPU 1 has a total capacity of 139.81 GiB of which 135.02 GiB is free. Including non-PyTorch memory, this process has 4.78 GiB memory in use. Of the allocated memory 3.24 GiB is allocated by PyTorch, and 65.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank10]: Traceback (most recent call last): +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank10]: pretrain( +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank10]: iteration, num_floating_point_operations_so_far = train( +[rank10]: ^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank10]: ) = train_step( +[rank10]: ^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank10]: losses_reduced = forward_backward_func( +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank10]: output_tensor, num_tokens = forward_step( +[rank10]: ^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank10]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank10]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank10]: batch = next(global_batches) +[rank10]: ^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank10]: attention_mask = torch.ones( +[rank10]: ^^^^^^^^^^^ +[rank10]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 4096.00 GiB. GPU 2 has a total capacity of 139.81 GiB of which 135.00 GiB is free. Including non-PyTorch memory, this process has 4.80 GiB memory in use. Of the allocated memory 3.24 GiB is allocated by PyTorch, and 65.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank14]: Traceback (most recent call last): +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank14]: pretrain( +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank14]: iteration, num_floating_point_operations_so_far = train( +[rank14]: ^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank14]: ) = train_step( +[rank14]: ^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank14]: losses_reduced = forward_backward_func( +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank14]: output_tensor, num_tokens = forward_step( +[rank14]: ^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank14]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank14]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank14]: batch = next(global_batches) +[rank14]: ^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank14]: attention_mask = torch.ones( +[rank14]: ^^^^^^^^^^^ +[rank14]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 4096.00 GiB. GPU 6 has a total capacity of 139.81 GiB of which 135.00 GiB is free. Including non-PyTorch memory, this process has 4.80 GiB memory in use. Of the allocated memory 3.24 GiB is allocated by PyTorch, and 65.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank12]: Traceback (most recent call last): +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank12]: pretrain( +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank12]: iteration, num_floating_point_operations_so_far = train( +[rank12]: ^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank12]: ) = train_step( +[rank12]: ^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank12]: losses_reduced = forward_backward_func( +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank12]: output_tensor, num_tokens = forward_step( +[rank12]: ^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank12]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank12]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: Traceback (most recent call last): +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank1]: pretrain( +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank1]: iteration, num_floating_point_operations_so_far = train( +[rank1]: ^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank1]: ) = train_step( +[rank1]: ^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank1]: losses_reduced = forward_backward_func( +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank12]: batch = next(global_batches) +[rank12]: ^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank12]: attention_mask = torch.ones( +[rank12]: ^^^^^^^^^^^ +[rank12]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 4096.00 GiB. GPU 4 has a total capacity of 139.81 GiB of which 135.00 GiB is free. Including non-PyTorch memory, this process has 4.80 GiB memory in use. Of the allocated memory 3.24 GiB is allocated by PyTorch, and 65.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank1]: output_tensor, num_tokens = forward_step( +[rank1]: ^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank1]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank1]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank1]: batch = next(global_batches) +[rank1]: ^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank1]: attention_mask = torch.ones( +[rank1]: ^^^^^^^^^^^ +[rank1]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 4096.00 GiB. GPU 1 has a total capacity of 139.81 GiB of which 135.00 GiB is free. Including non-PyTorch memory, this process has 4.80 GiB memory in use. Of the allocated memory 3.24 GiB is allocated by PyTorch, and 65.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank4]: Traceback (most recent call last): +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank4]: pretrain( +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank4]: iteration, num_floating_point_operations_so_far = train( +[rank4]: ^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank4]: ) = train_step( +[rank4]: ^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank4]: losses_reduced = forward_backward_func( +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank4]: output_tensor, num_tokens = forward_step( +[rank4]: ^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank4]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank4]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank4]: batch = next(global_batches) +[rank4]: ^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank4]: attention_mask = torch.ones( +[rank4]: ^^^^^^^^^^^ +[rank4]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 4096.00 GiB. GPU 4 has a total capacity of 139.81 GiB of which 135.02 GiB is free. Including non-PyTorch memory, this process has 4.78 GiB memory in use. Of the allocated memory 3.24 GiB is allocated by PyTorch, and 65.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank7]: Traceback (most recent call last): +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank7]: pretrain( +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank7]: iteration, num_floating_point_operations_so_far = train( +[rank7]: ^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank7]: ) = train_step( +[rank7]: ^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank7]: losses_reduced = forward_backward_func( +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank7]: output_tensor, num_tokens = forward_step( +[rank7]: ^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank7]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank7]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank7]: batch = next(global_batches) +[rank7]: ^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank7]: attention_mask = torch.ones( +[rank7]: ^^^^^^^^^^^ +[rank7]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 4096.00 GiB. GPU 7 has a total capacity of 139.81 GiB of which 135.00 GiB is free. Including non-PyTorch memory, this process has 4.80 GiB memory in use. Of the allocated memory 3.24 GiB is allocated by PyTorch, and 65.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank5]: Traceback (most recent call last): +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank5]: pretrain( +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank5]: iteration, num_floating_point_operations_so_far = train( +[rank5]: ^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank5]: ) = train_step( +[rank5]: ^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank5]: losses_reduced = forward_backward_func( +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank5]: output_tensor, num_tokens = forward_step( +[rank5]: ^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank5]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank5]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank5]: batch = next(global_batches) +[rank5]: ^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank5]: attention_mask = torch.ones( +[rank5]: ^^^^^^^^^^^ +[rank5]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 4096.00 GiB. GPU 5 has a total capacity of 139.81 GiB of which 135.00 GiB is free. Including non-PyTorch memory, this process has 4.80 GiB memory in use. Of the allocated memory 3.24 GiB is allocated by PyTorch, and 65.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank0]: Traceback (most recent call last): +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank0]: pretrain( +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank0]: iteration, num_floating_point_operations_so_far = train( +[rank0]: ^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank0]: ) = train_step( +[rank0]: ^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank0]: losses_reduced = forward_backward_func( +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank0]: output_tensor, num_tokens = forward_step( +[rank0]: ^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank0]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank0]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank0]: batch = next(global_batches) +[rank0]: ^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank0]: attention_mask = torch.ones( +[rank0]: ^^^^^^^^^^^ +[rank0]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 4096.00 GiB. GPU 0 has a total capacity of 139.81 GiB of which 135.02 GiB is free. Including non-PyTorch memory, this process has 4.78 GiB memory in use. Of the allocated memory 3.24 GiB is allocated by PyTorch, and 65.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank2]: Traceback (most recent call last): +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank2]: pretrain( +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank2]: iteration, num_floating_point_operations_so_far = train( +[rank2]: ^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank2]: ) = train_step( +[rank2]: ^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank2]: losses_reduced = forward_backward_func( +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank2]: output_tensor, num_tokens = forward_step( +[rank2]: ^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank2]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank2]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank2]: batch = next(global_batches) +[rank2]: ^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank2]: attention_mask = torch.ones( +[rank2]: ^^^^^^^^^^^ +[rank2]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 4096.00 GiB. GPU 2 has a total capacity of 139.81 GiB of which 135.02 GiB is free. Including non-PyTorch memory, this process has 4.78 GiB memory in use. Of the allocated memory 3.24 GiB is allocated by PyTorch, and 65.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank6]: Traceback (most recent call last): +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank6]: pretrain( +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank6]: iteration, num_floating_point_operations_so_far = train( +[rank6]: ^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank6]: ) = train_step( +[rank6]: ^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank6]: losses_reduced = forward_backward_func( +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank6]: output_tensor, num_tokens = forward_step( +[rank6]: ^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank6]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank6]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank6]: batch = next(global_batches) +[rank6]: ^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank6]: attention_mask = torch.ones( +[rank6]: ^^^^^^^^^^^ +[rank6]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 4096.00 GiB. GPU 6 has a total capacity of 139.81 GiB of which 135.02 GiB is free. Including non-PyTorch memory, this process has 4.78 GiB memory in use. Of the allocated memory 3.24 GiB is allocated by PyTorch, and 65.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank1]:[W621 22:18:04.713630731 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank11]:[W621 22:18:04.874247250 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank9]:[W621 22:18:04.884371913 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank3]:[W621 22:18:04.755348522 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank7]:[W621 22:18:04.765307811 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank15]:[W621 22:18:04.914875220 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank13]:[W621 22:18:04.914989293 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank5]:[W621 22:18:05.047467593 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +W0621 22:18:06.355000 1086332 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1086404 closing signal SIGTERM +W0621 22:18:06.357000 1086332 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1086405 closing signal SIGTERM +W0621 22:18:06.358000 1086332 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1086406 closing signal SIGTERM +W0621 22:18:06.380000 1086332 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1086408 closing signal SIGTERM +W0621 22:18:06.384000 1086332 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1086409 closing signal SIGTERM +W0621 22:18:06.385000 1086332 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1086410 closing signal SIGTERM +W0621 22:18:06.386000 1086332 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1086411 closing signal SIGTERM +W0621 22:18:06.421000 2009444 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2009513 closing signal SIGTERM +W0621 22:18:06.424000 2009444 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2009515 closing signal SIGTERM +W0621 22:18:06.427000 2009444 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2009517 closing signal SIGTERM +W0621 22:18:06.430000 2009444 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2009519 closing signal SIGTERM +W0621 22:18:06.433000 2009444 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2009520 closing signal SIGTERM +E0621 22:18:06.645000 2009444 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 1 (pid: 2009514) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +Traceback (most recent call last): + File "", line 198, in _run_module_as_main + File "", line 88, in _run_code + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in + main() + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper + return arg(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main + launch(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch + run(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run + elastic_launch( + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__ + return launch_agent(self._config, self._entrypoint, list(args)) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent + raise ChildFailedError( +torch.distributed.elastic.multiprocessing.errors.ChildFailedError: +============================================================ +./pretrain_gpt_profile.py FAILED +------------------------------------------------------------ +Failures: +[1]: + time : 2025-06-21_22:18:06 + host : fs-mbz-gpu-286 + rank : 11 (local_rank: 3) + exitcode : 1 (pid: 2009516) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +[2]: + time : 2025-06-21_22:18:06 + host : fs-mbz-gpu-286 + rank : 13 (local_rank: 5) + exitcode : 1 (pid: 2009518) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +------------------------------------------------------------ +Root Cause (first observed failure): +[0]: + time : 2025-06-21_22:18:06 + host : fs-mbz-gpu-286 + rank : 9 (local_rank: 1) + exitcode : 1 (pid: 2009514) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +============================================================ +E0621 22:18:06.746000 1086332 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 3 (pid: 1086407) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +Traceback (most recent call last): + File "", line 198, in _run_module_as_main + File "", line 88, in _run_code + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in + main() + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper + return arg(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main + launch(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch + run(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run + elastic_launch( + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__ + return launch_agent(self._config, self._entrypoint, list(args)) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent + raise ChildFailedError( +torch.distributed.elastic.multiprocessing.errors.ChildFailedError: +============================================================ +./pretrain_gpt_profile.py FAILED +------------------------------------------------------------ +Failures: + +------------------------------------------------------------ +Root Cause (first observed failure): +[0]: + time : 2025-06-21_22:18:06 + host : fs-mbz-gpu-239 + rank : 3 (local_rank: 3) + exitcode : 1 (pid: 1086407) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +============================================================ ++ set +x ++ set +x ++ for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072 ++ export PROF_CTX_LENGTH=40960 ++ PROF_CTX_LENGTH=40960 ++ name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L40960*tp2.cp8.bs16.json' ++ '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L40960*tp2.cp8.bs16.json' ']' ++ echo 'Running ctx_length=40960, TP_SIZE=2, CP_SIZE=8, BATCH_SIZE=16' ++ srun bash ./attnserver.sh ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 1 --rdzv_id 343241 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-239:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 40960 --max-position-embeddings 40960 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 0 --rdzv_id 343241 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-239:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 40960 --max-position-embeddings 40960 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:18:11.149000 2011243 site-packages/torch/distributed/run.py:766] +W0621 22:18:11.149000 2011243 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:18:11.149000 2011243 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:18:11.149000 2011243 site-packages/torch/distributed/run.py:766] ***************************************** +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:18:11.406000 1088184 site-packages/torch/distributed/run.py:766] +W0621 22:18:11.406000 1088184 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:18:11.406000 1088184 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:18:11.406000 1088184 site-packages/torch/distributed/run.py:766] ***************************************** +[rank2]:[W621 22:18:34.721045525 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 2] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank13]:[W621 22:18:34.860611110 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 13] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank6]:[W621 22:18:34.721239130 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 6] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank4]:[W621 22:18:34.721250280 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 4] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank14]:[W621 22:18:34.860946411 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 14] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank5]:[W621 22:18:34.727510509 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 5] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank10]:[W621 22:18:34.861016969 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 10] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank9]:[W621 22:18:34.861043933 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 9] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank7]:[W621 22:18:34.727788113 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 7] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank15]:[W621 22:18:34.861211969 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 15] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank3]:[W621 22:18:34.728018685 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 3] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank11]:[W621 22:18:34.861253566 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 11] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank12]:[W621 22:18:34.863124569 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 12] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank1]:[W621 22:18:34.729028300 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 1] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank8]:[W621 22:18:34.943613022 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 8] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank0]:[W621 22:18:34.870233078 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 0] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/models/gpt/gpt_layer_specs.py:94: UserWarning: The fp8 argument in "get_gpt_layer_with_transformer_engine_spec" has been deprecated and will be removed soon. Please update your code accordingly. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/transformer_engine/pytorch/cpu_offload.py:595: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +[rank11]: Traceback (most recent call last): +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank11]: pretrain( +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank11]: iteration, num_floating_point_operations_so_far = train( +[rank11]: ^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank11]: ) = train_step( +[rank11]: ^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank11]: losses_reduced = forward_backward_func( +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank11]: output_tensor, num_tokens = forward_step( +[rank11]: ^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank11]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank11]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank11]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank11]: batch = next(global_batches) +[rank11]: ^^^^^^^^^^^^^^^^^^^^ +[rank11]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank11]: attention_mask = torch.ones( +[rank11]: ^^^^^^^^^^^ +[rank11]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 6400.00 GiB. GPU 3 has a total capacity of 139.81 GiB of which 134.69 GiB is free. Including non-PyTorch memory, this process has 5.12 GiB memory in use. Of the allocated memory 3.53 GiB is allocated by PyTorch, and 110.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank9]: Traceback (most recent call last): +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank9]: pretrain( +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank9]: iteration, num_floating_point_operations_so_far = train( +[rank9]: ^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank9]: ) = train_step( +[rank9]: ^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank9]: losses_reduced = forward_backward_func( +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank9]: output_tensor, num_tokens = forward_step( +[rank9]: ^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank9]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank9]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank9]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank9]: batch = next(global_batches) +[rank9]: ^^^^^^^^^^^^^^^^^^^^ +[rank9]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank9]: attention_mask = torch.ones( +[rank9]: ^^^^^^^^^^^ +[rank9]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 6400.00 GiB. GPU 1 has a total capacity of 139.81 GiB of which 134.69 GiB is free. Including non-PyTorch memory, this process has 5.12 GiB memory in use. Of the allocated memory 3.53 GiB is allocated by PyTorch, and 110.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank8]: Traceback (most recent call last): +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank8]: pretrain( +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank8]: iteration, num_floating_point_operations_so_far = train( +[rank8]: ^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank8]: ) = train_step( +[rank8]: ^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank8]: losses_reduced = forward_backward_func( +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank8]: output_tensor, num_tokens = forward_step( +[rank8]: ^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank8]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank8]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank8]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank8]: batch = next(global_batches) +[rank8]: ^^^^^^^^^^^^^^^^^^^^ +[rank8]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank8]: attention_mask = torch.ones( +[rank8]: ^^^^^^^^^^^ +[rank8]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 6400.00 GiB. GPU 0 has a total capacity of 139.81 GiB of which 134.67 GiB is free. Including non-PyTorch memory, this process has 5.13 GiB memory in use. Of the allocated memory 3.53 GiB is allocated by PyTorch, and 110.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank15]: Traceback (most recent call last): +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank15]: pretrain( +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank15]: iteration, num_floating_point_operations_so_far = train( +[rank15]: ^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank15]: ) = train_step( +[rank15]: ^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank15]: losses_reduced = forward_backward_func( +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank15]: output_tensor, num_tokens = forward_step( +[rank15]: ^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank15]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank15]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank15]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank15]: batch = next(global_batches) +[rank15]: ^^^^^^^^^^^^^^^^^^^^ +[rank15]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank15]: attention_mask = torch.ones( +[rank15]: ^^^^^^^^^^^ +[rank15]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 6400.00 GiB. GPU 7 has a total capacity of 139.81 GiB of which 134.69 GiB is free. Including non-PyTorch memory, this process has 5.12 GiB memory in use. Of the allocated memory 3.53 GiB is allocated by PyTorch, and 110.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank10]: Traceback (most recent call last): +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank10]: pretrain( +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank10]: iteration, num_floating_point_operations_so_far = train( +[rank10]: ^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank10]: ) = train_step( +[rank10]: ^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank10]: losses_reduced = forward_backward_func( +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank10]: output_tensor, num_tokens = forward_step( +[rank10]: ^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank10]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank10]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank10]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank10]: batch = next(global_batches) +[rank10]: ^^^^^^^^^^^^^^^^^^^^ +[rank10]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank10]: attention_mask = torch.ones( +[rank10]: ^^^^^^^^^^^ +[rank10]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 6400.00 GiB. GPU 2 has a total capacity of 139.81 GiB of which 134.67 GiB is free. Including non-PyTorch memory, this process has 5.13 GiB memory in use. Of the allocated memory 3.53 GiB is allocated by PyTorch, and 110.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank12]: Traceback (most recent call last): +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank12]: pretrain( +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank12]: iteration, num_floating_point_operations_so_far = train( +[rank12]: ^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank12]: ) = train_step( +[rank12]: ^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank12]: losses_reduced = forward_backward_func( +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank12]: output_tensor, num_tokens = forward_step( +[rank12]: ^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank12]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank12]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank12]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank12]: batch = next(global_batches) +[rank12]: ^^^^^^^^^^^^^^^^^^^^ +[rank12]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank12]: attention_mask = torch.ones( +[rank12]: ^^^^^^^^^^^ +[rank12]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 6400.00 GiB. GPU 4 has a total capacity of 139.81 GiB of which 134.67 GiB is free. Including non-PyTorch memory, this process has 5.13 GiB memory in use. Of the allocated memory 3.53 GiB is allocated by PyTorch, and 110.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank13]: Traceback (most recent call last): +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank13]: pretrain( +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank13]: iteration, num_floating_point_operations_so_far = train( +[rank13]: ^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank13]: ) = train_step( +[rank13]: ^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank13]: losses_reduced = forward_backward_func( +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank13]: output_tensor, num_tokens = forward_step( +[rank13]: ^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank13]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank13]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank13]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank13]: batch = next(global_batches) +[rank13]: ^^^^^^^^^^^^^^^^^^^^ +[rank13]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank13]: attention_mask = torch.ones( +[rank13]: ^^^^^^^^^^^ +[rank13]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 6400.00 GiB. GPU 5 has a total capacity of 139.81 GiB of which 134.69 GiB is free. Including non-PyTorch memory, this process has 5.12 GiB memory in use. Of the allocated memory 3.53 GiB is allocated by PyTorch, and 110.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank14]: Traceback (most recent call last): +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank14]: pretrain( +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank14]: iteration, num_floating_point_operations_so_far = train( +[rank14]: ^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank14]: ) = train_step( +[rank14]: ^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank14]: losses_reduced = forward_backward_func( +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank14]: output_tensor, num_tokens = forward_step( +[rank14]: ^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank14]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank14]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank14]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank14]: batch = next(global_batches) +[rank14]: ^^^^^^^^^^^^^^^^^^^^ +[rank14]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank14]: attention_mask = torch.ones( +[rank14]: ^^^^^^^^^^^ +[rank14]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 6400.00 GiB. GPU 6 has a total capacity of 139.81 GiB of which 134.67 GiB is free. Including non-PyTorch memory, this process has 5.13 GiB memory in use. Of the allocated memory 3.53 GiB is allocated by PyTorch, and 110.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank3]: Traceback (most recent call last): +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank3]: pretrain( +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank3]: iteration, num_floating_point_operations_so_far = train( +[rank3]: ^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank3]: ) = train_step( +[rank3]: ^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank3]: losses_reduced = forward_backward_func( +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank3]: output_tensor, num_tokens = forward_step( +[rank3]: ^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank3]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank3]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank3]: batch = next(global_batches) +[rank3]: ^^^^^^^^^^^^^^^^^^^^ +[rank3]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank3]: attention_mask = torch.ones( +[rank3]: ^^^^^^^^^^^ +[rank3]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 6400.00 GiB. GPU 3 has a total capacity of 139.81 GiB of which 134.67 GiB is free. Including non-PyTorch memory, this process has 5.13 GiB memory in use. Of the allocated memory 3.53 GiB is allocated by PyTorch, and 110.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank7]: Traceback (most recent call last): +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank7]: pretrain( +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank7]: iteration, num_floating_point_operations_so_far = train( +[rank7]: ^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank7]: ) = train_step( +[rank7]: ^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank7]: losses_reduced = forward_backward_func( +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank7]: output_tensor, num_tokens = forward_step( +[rank7]: ^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank7]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank7]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank7]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank7]: batch = next(global_batches) +[rank7]: ^^^^^^^^^^^^^^^^^^^^ +[rank7]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank7]: attention_mask = torch.ones( +[rank7]: ^^^^^^^^^^^ +[rank7]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 6400.00 GiB. GPU 7 has a total capacity of 139.81 GiB of which 134.67 GiB is free. Including non-PyTorch memory, this process has 5.13 GiB memory in use. Of the allocated memory 3.53 GiB is allocated by PyTorch, and 110.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank4]: Traceback (most recent call last): +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank4]: pretrain( +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank4]: iteration, num_floating_point_operations_so_far = train( +[rank4]: ^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank4]: ) = train_step( +[rank4]: ^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank4]: losses_reduced = forward_backward_func( +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank4]: output_tensor, num_tokens = forward_step( +[rank4]: ^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank4]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank4]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank4]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank4]: batch = next(global_batches) +[rank4]: ^^^^^^^^^^^^^^^^^^^^ +[rank4]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank4]: attention_mask = torch.ones( +[rank4]: ^^^^^^^^^^^ +[rank4]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 6400.00 GiB. GPU 4 has a total capacity of 139.81 GiB of which 134.69 GiB is free. Including non-PyTorch memory, this process has 5.12 GiB memory in use. Of the allocated memory 3.53 GiB is allocated by PyTorch, and 110.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank0]: Traceback (most recent call last): +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank0]: pretrain( +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank0]: iteration, num_floating_point_operations_so_far = train( +[rank0]: ^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank0]: ) = train_step( +[rank0]: ^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank0]: losses_reduced = forward_backward_func( +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank0]: output_tensor, num_tokens = forward_step( +[rank0]: ^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank0]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank0]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank0]: batch = next(global_batches) +[rank0]: ^^^^^^^^^^^^^^^^^^^^ +[rank0]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank0]: attention_mask = torch.ones( +[rank0]: ^^^^^^^^^^^ +[rank0]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 6400.00 GiB. GPU 0 has a total capacity of 139.81 GiB of which 134.69 GiB is free. Including non-PyTorch memory, this process has 5.12 GiB memory in use. Of the allocated memory 3.53 GiB is allocated by PyTorch, and 110.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank1]: Traceback (most recent call last): +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank1]: pretrain( +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank1]: iteration, num_floating_point_operations_so_far = train( +[rank1]: ^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank1]: ) = train_step( +[rank1]: ^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank1]: losses_reduced = forward_backward_func( +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank1]: output_tensor, num_tokens = forward_step( +[rank1]: ^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank1]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank1]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank1]: batch = next(global_batches) +[rank1]: ^^^^^^^^^^^^^^^^^^^^ +[rank1]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank1]: attention_mask = torch.ones( +[rank1]: ^^^^^^^^^^^ +[rank1]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 6400.00 GiB. GPU 1 has a total capacity of 139.81 GiB of which 134.67 GiB is free. Including non-PyTorch memory, this process has 5.13 GiB memory in use. Of the allocated memory 3.53 GiB is allocated by PyTorch, and 110.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank5]: Traceback (most recent call last): +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank5]: pretrain( +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank5]: iteration, num_floating_point_operations_so_far = train( +[rank5]: ^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank5]: ) = train_step( +[rank5]: ^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank5]: losses_reduced = forward_backward_func( +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank5]: output_tensor, num_tokens = forward_step( +[rank5]: ^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank5]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank5]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank5]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank5]: batch = next(global_batches) +[rank5]: ^^^^^^^^^^^^^^^^^^^^ +[rank5]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank5]: attention_mask = torch.ones( +[rank5]: ^^^^^^^^^^^ +[rank5]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 6400.00 GiB. GPU 5 has a total capacity of 139.81 GiB of which 134.67 GiB is free. Including non-PyTorch memory, this process has 5.13 GiB memory in use. Of the allocated memory 3.53 GiB is allocated by PyTorch, and 110.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank2]: Traceback (most recent call last): +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank2]: pretrain( +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank2]: iteration, num_floating_point_operations_so_far = train( +[rank2]: ^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank2]: ) = train_step( +[rank2]: ^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank2]: losses_reduced = forward_backward_func( +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank2]: output_tensor, num_tokens = forward_step( +[rank2]: ^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank2]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank2]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank2]: batch = next(global_batches) +[rank2]: ^^^^^^^^^^^^^^^^^^^^ +[rank2]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank2]: attention_mask = torch.ones( +[rank2]: ^^^^^^^^^^^ +[rank2]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 6400.00 GiB. GPU 2 has a total capacity of 139.81 GiB of which 134.69 GiB is free. Including non-PyTorch memory, this process has 5.12 GiB memory in use. Of the allocated memory 3.53 GiB is allocated by PyTorch, and 110.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank6]: Traceback (most recent call last): +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 554, in +[rank6]: pretrain( +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 863, in pretrain +[rank6]: iteration, num_floating_point_operations_so_far = train( +[rank6]: ^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 2229, in train +[rank6]: ) = train_step( +[rank6]: ^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/training/training.py", line 1382, in train_step +[rank6]: losses_reduced = forward_backward_func( +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 518, in forward_backward_no_pipelining +[rank6]: output_tensor, num_tokens = forward_step( +[rank6]: ^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/megatron/core/pipeline_parallel/schedules.py", line 289, in forward_step +[rank6]: output_tensor, loss_func = forward_step_func(data_iterator, model) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 446, in forward_step +[rank6]: (tokens, labels, loss_mask, attention_mask, position_ids), token_lens = get_batch(data_iterator) +[rank6]: ^^^^^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 284, in get_batch +[rank6]: batch = next(global_batches) +[rank6]: ^^^^^^^^^^^^^^^^^^^^ +[rank6]: File "/mnt/weka/home/hao.zhang/junda/attnserver-megatron/./pretrain_gpt_profile.py", line 226, in setup_batches +[rank6]: attention_mask = torch.ones( +[rank6]: ^^^^^^^^^^^ +[rank6]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 6400.00 GiB. GPU 6 has a total capacity of 139.81 GiB of which 134.69 GiB is free. Including non-PyTorch memory, this process has 5.12 GiB memory in use. Of the allocated memory 3.53 GiB is allocated by PyTorch, and 110.54 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) +[rank11]:[W621 22:18:45.828084593 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank3]:[W621 22:18:46.082246821 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank1]:[W621 22:18:46.088734107 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank7]:[W621 22:18:46.092369870 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank13]:[W621 22:18:46.241916677 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank9]:[W621 22:18:46.262139016 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank5]:[W621 22:18:46.133047519 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +[rank15]:[W621 22:18:46.272498224 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) +W0621 22:18:47.246000 2011243 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2011313 closing signal SIGTERM +W0621 22:18:47.252000 2011243 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2011314 closing signal SIGTERM +W0621 22:18:47.253000 2011243 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2011315 closing signal SIGTERM +W0621 22:18:47.255000 2011243 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2011317 closing signal SIGTERM +W0621 22:18:47.258000 2011243 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2011318 closing signal SIGTERM +W0621 22:18:47.259000 2011243 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2011319 closing signal SIGTERM +W0621 22:18:47.261000 2011243 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2011320 closing signal SIGTERM +W0621 22:18:47.747000 1088184 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1088256 closing signal SIGTERM +W0621 22:18:47.750000 1088184 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1088258 closing signal SIGTERM +W0621 22:18:47.753000 1088184 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1088260 closing signal SIGTERM +W0621 22:18:47.757000 1088184 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1088261 closing signal SIGTERM +W0621 22:18:47.758000 1088184 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 1088262 closing signal SIGTERM +E0621 22:18:47.874000 2011243 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 3 (pid: 2011316) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 +Traceback (most recent call last): + File "", line 198, in _run_module_as_main + File "", line 88, in _run_code + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in + main() + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper + return arg(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main + launch(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch + run(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run + elastic_launch( + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__ + return launch_agent(self._config, self._entrypoint, list(args)) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +E0621 22:18:47.908000 1088184 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 1 (pid: 1088257) of binary: /mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/bin/python3 + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent + raise ChildFailedError( +torch.distributed.elastic.multiprocessing.errors.ChildFailedError: +============================================================ +./pretrain_gpt_profile.py FAILED +------------------------------------------------------------ +Failures: + +------------------------------------------------------------ +Root Cause (first observed failure): +[0]: + time : 2025-06-21_22:18:47 + host : fs-mbz-gpu-286 + rank : 11 (local_rank: 3) + exitcode : 1 (pid: 2011316) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +============================================================ +Traceback (most recent call last): + File "", line 198, in _run_module_as_main + File "", line 88, in _run_code + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 207, in + main() + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/typing_extensions.py", line 3253, in wrapper + return arg(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 203, in main + launch(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py", line 188, in launch + run(args) + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run + elastic_launch( + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__ + return launch_agent(self._config, self._entrypoint, list(args)) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent + raise ChildFailedError( +torch.distributed.elastic.multiprocessing.errors.ChildFailedError: +============================================================ +./pretrain_gpt_profile.py FAILED +------------------------------------------------------------ +Failures: +[1]: + time : 2025-06-21_22:18:47 + host : fs-mbz-gpu-239 + rank : 3 (local_rank: 3) + exitcode : 1 (pid: 1088259) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +[2]: + time : 2025-06-21_22:18:47 + host : fs-mbz-gpu-239 + rank : 7 (local_rank: 7) + exitcode : 1 (pid: 1088263) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +------------------------------------------------------------ +Root Cause (first observed failure): +[0]: + time : 2025-06-21_22:18:47 + host : fs-mbz-gpu-239 + rank : 1 (local_rank: 1) + exitcode : 1 (pid: 1088257) + error_file: + traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html +============================================================ ++ set +x ++ set +x ++ for ctx_length in 1024 2048 4096 8192 12288 16384 24576 32768 40960 49152 65536 81920 98304 131072 ++ export PROF_CTX_LENGTH=49152 ++ PROF_CTX_LENGTH=49152 ++ name='/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L49152*tp2.cp8.bs16.json' ++ '[' -f '/mnt/sharefs/users/hao.zhang/junda/megatron-prof-data--unstable-v5/mytrace.L49152*tp2.cp8.bs16.json' ']' ++ echo 'Running ctx_length=49152, TP_SIZE=2, CP_SIZE=8, BATCH_SIZE=16' ++ srun bash ./attnserver.sh ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 0 --rdzv_id 343241 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-239:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 49152 --max-position-embeddings 49152 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ ++ which python3 ++ python3 -m torch.distributed.launch --nproc_per_node 8 --nnodes 2 --node_rank 1 --rdzv_id 343241 --rdzv_backend c10d --rdzv_endpoint fs-mbz-gpu-239:29500 ./pretrain_gpt_profile.py --tensor-model-parallel-size 2 --context-parallel-size 8 --num-layers 2 --hidden-size 4096 --num-attention-heads 64 --group-query-attention --num-query-groups 16 --seq-length 49152 --max-position-embeddings 49152 --micro-batch-size 1 --global-batch-size 1 --lr 0.0005 --train-iters 10 --lr-decay-iters 150000 --lr-decay-style cosine --lr-warmup-iters 2 --weight-decay .1 --adam-beta2 .999 --fp16 --log-interval 1 --save-interval 16 --eval-interval 16 --eval-iters 1 --vocab-file vocab.json --merge-file merges.txt --save gpt-checkpoint --load gpt-checkpoint --logging-level 0 --mock-data --tensorboard-dir tensorboard-logs/ +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:18:52.542000 1090052 site-packages/torch/distributed/run.py:766] +W0621 22:18:52.542000 1090052 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:18:52.542000 1090052 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:18:52.542000 1090052 site-packages/torch/distributed/run.py:766] ***************************************** +/mnt/weka/home/hao.zhang/conda/miniconda/envs/junda-attnserver/lib/python3.12/site-packages/torch/distributed/launch.py:207: FutureWarning: The module torch.distributed.launch is deprecated +and will be removed in future. Use torchrun. +Note that --use-env is set by default in torchrun. +If your script expects `--local-rank` argument to be set, please +change it to read from `os.environ['LOCAL_RANK']` instead. See +https://pytorch.org/docs/stable/distributed.html#launch-utility for +further instructions + + main() +W0621 22:18:52.548000 2013060 site-packages/torch/distributed/run.py:766] +W0621 22:18:52.548000 2013060 site-packages/torch/distributed/run.py:766] ***************************************** +W0621 22:18:52.548000 2013060 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0621 22:18:52.548000 2013060 site-packages/torch/distributed/run.py:766] ***************************************** +[rank3]:[W621 22:19:15.636354897 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 3] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank1]:[W621 22:19:15.636413421 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 1] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank7]:[W621 22:19:15.636428932 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 7] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank11]:[W621 22:19:15.767881970 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 11] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank13]:[W621 22:19:15.767881857 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 13] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank15]:[W621 22:19:15.767881899 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 15] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank9]:[W621 22:19:15.767929244 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 9] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank5]:[W621 22:19:15.643217876 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 5] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank2]:[W621 22:19:15.644819460 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 2] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank6]:[W621 22:19:15.647484641 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 6] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank4]:[W621 22:19:15.647556266 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 4] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank12]:[W621 22:19:15.790237214 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 12] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank10]:[W621 22:19:15.790290069 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 10] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank14]:[W621 22:19:15.790555126 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 14] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank8]:[W621 22:19:15.873168372 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 8] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device. +[rank0]:[W621 22:19:15.785074172 ProcessGroupNCCL.cpp:4715] [PG ID 0 PG GUID 0 Rank 0] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can pecify device_id in init_process_group() to force use of a particular device.