waveydaveygravy/c10wnfac3-v3

8 instance images. 120 repeats 1 epocj batchsize 1

trigger word c10wnfac3

command:

• accelerate
• launch
• --config_file=/content/kohya-trainer/accelerate_config/config.yaml
• --num_cpu_threads_per_process=8
• train_network.py
• --network_dim=128
• --network_alpha=128
• --network_module=networks.lora
• --learning_rate=0.0001
• --text_encoder_lr=5e-05
• --training_comment=this_collab_is_RAGE_INDUCING
• --lr_scheduler=cosine
• --pretrained_model_name_or_path=/content/pre_trained_model/Stable-Diffusion-v1-5.ckpt
• --vae=/content/vae/stablediffusion.vae.pt
• --caption_extension=.txt
• --train_data_dir=/content/dreambooth/train_data
• --reg_data_dir=/content/dreambooth/reg_data
• --output_dir=/content/dreambooth/output
• --prior_loss_weight=1.0
• --output_name=c10wnfac3
• --mixed_precision=fp16
• --save_precision=fp16
• --save_every_n_epochs=1
• --save_model_as=safetensors
• --resolution=512
• --train_batch_size=1
• --max_token_length=225
• --use_8bit_adam
• --max_train_epochs=1
• --gradient_checkpointing
• --gradient_accumulation_steps=1
• --clip_skip=2
• --logging_dir=/content/dreambooth/logs
• --log_prefix=c10wnfac3
• --shuffle_caption
• --xformers