*.shards.stepNNNNNN/state.pt holds only the EMA payload (state['extra']['ema'], fp32, clean names) extracted from the original training checkpoint, fingerprint-verified against engine_meta.pt. Loadable directly by methods/lafm/vla_flow_ft.py:load_pretrain_trainables with init_use_ema=true. Optimizer states were not retained.1[
2 {
3 "run": "lafm_vla_flow_qwen2b_flow_official_Jul10",
4 "step_dir": "lafm_vla_flow_qwen2b_weights.ckpt.shards.step000300000",
5 "source": "ema",
6 "bytes": 1774673706
7 }
8]