microsoft/FrogNano-4B-2609
5B ā¢ Updated ā¢ 1.24k ā¢ 142
Carlo Moro PRO
cnmoro
AI & ML interests
Retrieval & Small Models & PTBR Resources
Recent Activity
liked a model 1 day ago
microsoft/FrogNano-4B-2609 liked a model 3 days ago
LiquidAI/d1-3B-GGUF liked a model 5 days ago
Cactus-Compute/whistleOrganizations
liked a
model 1 day ago
liked a
model 3 days ago
liked a
model 5 days ago
reacted to SeaWolf-AI's post with ā¤ļø 8 days ago
Post
6228
š» Data-center AI, now on a laptop: POCKET-Darwin-180B
We're releasing a 4-bit GGUF build of Darwin-180B-RSI, #1 on seven official Hugging Face leaderboards (self-reported), that runs without a GPU.
š¦ 360 GB ā 111 GB (4-bit GGUF, 4 files)
š„ļø No GPU: one server CPU (16 threads) at 18.4ā21.0 tokens/s
š» RTX 5060 laptop (8 GB VRAM) + 32 GB RAM: 4.17 tokens/s
š§ 128 GB mini PC: whole model in memory, no GPU needed
šÆ MMLU-Pro, 2,000 questions, paired: original 87.65% = 4-bit 87.65%
How?
Ā· Only ~3B of 180B parameters are active per token (10 of 512 experts)
Ā· llama.cpp streams just the needed experts from SSD, so 32 GB RAM is enough
Ā· Graft quantization: we took the proven Unsloth UD-Q4_K_XL base build and swapped in only the 300 tensors our RSI training changed (300/300 verified)
Under the hood is Model-level Recursive Self-Improvement. The model solves verifiable problems, keeps only its own solutions that check out as correct, and trains on them. No human-written solutions or reasoning traces.
Built for teams that can't send data to an external cloud (defense, finance, public sector) to run a top-tier model fully offline.
š Article: https://huggingface.co/blog/FINAL-Bench/data-center-ai-now-on-a-laptop-pocket-darwin-180b
š¤ Model: FINAL-Bench/POCKET-Darwin-180B-GGUF
š§¬ Original: FINAL-Bench/Darwin-180B-RSI
#Darwin #RSI #GGUF #llamacpp #OnDevice #MoE
We're releasing a 4-bit GGUF build of Darwin-180B-RSI, #1 on seven official Hugging Face leaderboards (self-reported), that runs without a GPU.
š¦ 360 GB ā 111 GB (4-bit GGUF, 4 files)
š„ļø No GPU: one server CPU (16 threads) at 18.4ā21.0 tokens/s
š» RTX 5060 laptop (8 GB VRAM) + 32 GB RAM: 4.17 tokens/s
š§ 128 GB mini PC: whole model in memory, no GPU needed
šÆ MMLU-Pro, 2,000 questions, paired: original 87.65% = 4-bit 87.65%
How?
Ā· Only ~3B of 180B parameters are active per token (10 of 512 experts)
Ā· llama.cpp streams just the needed experts from SSD, so 32 GB RAM is enough
Ā· Graft quantization: we took the proven Unsloth UD-Q4_K_XL base build and swapped in only the 300 tensors our RSI training changed (300/300 verified)
Under the hood is Model-level Recursive Self-Improvement. The model solves verifiable problems, keeps only its own solutions that check out as correct, and trains on them. No human-written solutions or reasoning traces.
Built for teams that can't send data to an external cloud (defense, finance, public sector) to run a top-tier model fully offline.
š Article: https://huggingface.co/blog/FINAL-Bench/data-center-ai-now-on-a-laptop-pocket-darwin-180b
š¤ Model: FINAL-Bench/POCKET-Darwin-180B-GGUF
š§¬ Original: FINAL-Bench/Darwin-180B-RSI
#Darwin #RSI #GGUF #llamacpp #OnDevice #MoE
- 30 replies
liked a
Space 11 days ago
liked 2
models 16 days ago
liked a
model 17 days ago
liked a
model 18 days ago
liked a
model 19 days ago
liked a
model 23 days ago
liked a
model 24 days ago
liked 2
models 26 days ago
liked a
model 28 days ago
liked 2
models about 1 month ago
upvoted a collection about 1 month ago
reacted to RiverRider's post with š about 1 month ago
Post
1788
A 339 KB linear probe on frozen features beats the fine-tuned baseline on ChestX-ray14.
Linear(5376, 14) on frozen google/gemma-4-31B-it hidden states. No fine-tuning, no radiology pretraining, no augmentation. All 112,120 images, official test_list.txt.
Wang et al. 2017, ResNet-50 fine-tuned end to end 0.7451
this probe, frozen backbone + linear head 0.7590
view-position only (shortcut baseline) 0.5896
shuffled labels (refit floor) 0.5002
Ahead on 12 of 14 findings.
The comparison is split-matched, and that took care to get right. The number everyone quotes, CheXNet's 0.8414, is on a different test set: their own random 70/10/20 partition, not the official list. Do not compare 0.7590 to it. The matched row is from Wang's v5 appendix, added specifically to report the published split. I had this wrong in our own code for a day, quoting a cross-split reference as a head-to-head, which is the error worth not repeating in public.
Three controls, because a bare AUROC here is not interpretable. Shuffled labels catch leakage. View-only catches the shortcut, since portable AP films are taken of sicker patients, and it is folded, because Hernia's raw view-only of 0.3436 is really 0.6564 of shortcut once flipped. Intervals resample patients and not images, since the test split is 25,596 films from 2,797 patients.
Banked negatives are on the card too. Max-pooling and top-16 pooling were predicted to help focal findings and did the opposite, costing 0.0537 and 0.0225. Readout depth barely matters, 0.7600 to 0.7605.
Scope: detection, not early detection. Research artifact, not a diagnostic device.
The backbone never runs in the demo. What ships is the reading.
Space: RiverRider/srt-cxr14-probe
Model: RiverRider/srt-cxr14-linear-probe
Data + states: RiverRider/srt-cxr14-frozen-probe
Linear(5376, 14) on frozen google/gemma-4-31B-it hidden states. No fine-tuning, no radiology pretraining, no augmentation. All 112,120 images, official test_list.txt.
Wang et al. 2017, ResNet-50 fine-tuned end to end 0.7451
this probe, frozen backbone + linear head 0.7590
view-position only (shortcut baseline) 0.5896
shuffled labels (refit floor) 0.5002
Ahead on 12 of 14 findings.
The comparison is split-matched, and that took care to get right. The number everyone quotes, CheXNet's 0.8414, is on a different test set: their own random 70/10/20 partition, not the official list. Do not compare 0.7590 to it. The matched row is from Wang's v5 appendix, added specifically to report the published split. I had this wrong in our own code for a day, quoting a cross-split reference as a head-to-head, which is the error worth not repeating in public.
Three controls, because a bare AUROC here is not interpretable. Shuffled labels catch leakage. View-only catches the shortcut, since portable AP films are taken of sicker patients, and it is folded, because Hernia's raw view-only of 0.3436 is really 0.6564 of shortcut once flipped. Intervals resample patients and not images, since the test split is 25,596 films from 2,797 patients.
Banked negatives are on the card too. Max-pooling and top-16 pooling were predicted to help focal findings and did the opposite, costing 0.0537 and 0.0225. Readout depth barely matters, 0.7600 to 0.7605.
Scope: detection, not early detection. Research artifact, not a diagnostic device.
The backbone never runs in the demo. What ships is the reading.
Space: RiverRider/srt-cxr14-probe
Model: RiverRider/srt-cxr14-linear-probe
Data + states: RiverRider/srt-cxr14-frozen-probe
- 3 replies
liked a
model about 1 month ago
