ckpt-countdown-families

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

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.

The two math-l5 runs v1 and v3 carry their periodic checkpoints on a 50-iteration grid. Their prune ran at the START of each dispatch, so the final dispatch of each was never pruned and left a dense every-ten tail of different length in each. Keeping multiples of 50 applies the same policy the runs applied to themselves earlier and makes the two comparable.

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
cd-gemma4b-fixed	google/gemma-3-4b-it	300	none
cd-gemma4b-fresh	google/gemma-3-4b-it	300	none
cd-gemma4b-hetero	google/gemma-3-4b-it	300	none
cd-gemma4b-mirror	google/gemma-3-4b-it	300	none
cd-gemma4b-mirror-v3	google/gemma-3-4b-it	300	none
cd-llama3b-fixed	meta-llama/Llama-3.2-3B-Instruct	300	none
cd-llama3b-fresh	meta-llama/Llama-3.2-3B-Instruct	300	none
cd-llama3b-hetero	meta-llama/Llama-3.2-3B-Instruct	300	none
cd-llama3b-mirror	meta-llama/Llama-3.2-3B-Instruct	300	none
cd-llama3b-mirror-v3	meta-llama/Llama-3.2-3B-Instruct	300	none
cd-llama8b-fixed	NousResearch/Meta-Llama-3.1-8B-Instruct	300	none
cd-llama8b-fresh	NousResearch/Meta-Llama-3.1-8B-Instruct	300	none
cd-llama8b-hetero	NousResearch/Meta-Llama-3.1-8B-Instruct	300	none
cd-llama8b-mirror	NousResearch/Meta-Llama-3.1-8B-Instruct	300	none
cd-llama8b-mirror-v3	NousResearch/Meta-Llama-3.1-8B-Instruct	300	none
cd-olmo7b-fixed	allenai/OLMo-2-1124-7B-Instruct	300	none
cd-olmo7b-fresh	allenai/OLMo-2-1124-7B-Instruct	300	none
cd-olmo7b-hetero	allenai/OLMo-2-1124-7B-Instruct	300	none
cd-olmo7b-mirror	allenai/OLMo-2-1124-7B-Instruct	300	none
cd-olmo7b-mirror-v3	allenai/OLMo-2-1124-7B-Instruct	300	none
cd-olmoe-fixed	allenai/OLMoE-1B-7B-0125-Instruct	300	none
cd-olmoe-fresh	allenai/OLMoE-1B-7B-0125-Instruct	300	none
cd-olmoe-hetero	allenai/OLMoE-1B-7B-0125-Instruct	300	none
cd-olmoe-mirror	allenai/OLMoE-1B-7B-0125-Instruct	300	none
cd-olmoe-mirror-v3	allenai/OLMoE-1B-7B-0125-Instruct	300	none