YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

ckpt-smallpop

Evolution-Strategies fine-tuning checkpoints from IBM Blue Vela.

The population-2 shaping grid on Countdown with Qwen2.5-3B, 1000 iterations per cell. The run name carries the shaping rule (z for z-scores, rh8 and rh30 for a running baseline at half-life 8 and 30) and k, where the learning rate is 1e-3/k.

iter<N>.pth are the periodic training checkpoints and final/pytorch_model.pth is the end-of-run save. The two are NOT copies of each other even at the same N -- the final save is written a few ES steps after the last periodic one, measured at a relative L2 difference of 0.002 to 0.004 where ten iterations move 0.008. Treat final/ as the canonical end-of-run weights.

One copy per iteration is published. Where a run has both a final save and a periodic checkpoint at that same iteration, only the final save is here. Periodic checkpoints at earlier iterations are kept, and a run that never reached its end keeps every periodic checkpoint it has.

Runs still training are not here yet. They are published once they reach their target and write a final save, so that no iteration ever appears twice.

All .pth are bf16 state dicts keyed by vLLM parameter names (fused qkv_proj, gate_up_proj), loadable by the trainer resume path.

run base model final iteration periodic checkpoints
cd3b-hetero-n2-rh30-k15 Qwen/Qwen2.5-3B-Instruct 1000 none
cd3b-hetero-n2-rh30-k2 Qwen/Qwen2.5-3B-Instruct 1000 none
cd3b-hetero-n2-rh30-k30 Qwen/Qwen2.5-3B-Instruct 1000 none
cd3b-hetero-n2-rh30-k4 Qwen/Qwen2.5-3B-Instruct 1000 none
cd3b-hetero-n2-rh30-k60 Qwen/Qwen2.5-3B-Instruct 1000 none
cd3b-hetero-n2-rh30-k8 Qwen/Qwen2.5-3B-Instruct 1000 none
cd3b-hetero-n2-rh8-k15 Qwen/Qwen2.5-3B-Instruct 1000 none
cd3b-hetero-n2-rh8-k2 Qwen/Qwen2.5-3B-Instruct 1000 none
cd3b-hetero-n2-rh8-k30 Qwen/Qwen2.5-3B-Instruct 1000 none
cd3b-hetero-n2-rh8-k4 Qwen/Qwen2.5-3B-Instruct 1000 none
cd3b-hetero-n2-rh8-k60 Qwen/Qwen2.5-3B-Instruct 1000 none
cd3b-hetero-n2-rh8-k8 Qwen/Qwen2.5-3B-Instruct 1000 none
cd3b-hetero-n2-z-k15 Qwen/Qwen2.5-3B-Instruct 1000 none
cd3b-hetero-n2-z-k2 Qwen/Qwen2.5-3B-Instruct 1000 none
cd3b-hetero-n2-z-k30 Qwen/Qwen2.5-3B-Instruct 1000 none
cd3b-hetero-n2-z-k4 Qwen/Qwen2.5-3B-Instruct 1000 none
cd3b-hetero-n2-z-k60 Qwen/Qwen2.5-3B-Instruct 1000 none
cd3b-hetero-n2-z-k8 Qwen/Qwen2.5-3B-Instruct 1000 none
Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support