weepiess2383/arc-layermap-flft-0813a

Archive: layermap-flft-0813a

EMA-consolidated research checkpoint archive.

Full-layer (no layer-map) LIBERO m323 finetune, 0813a. Control arm for the layer-map / first-28 frozen comparison.

Each *.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.

Manifest

[
  {
    "run": "lafm_vla_flow_neo2b_fulllayer_libero_m323_flft_0813a",
    "step_dir": "lafm_vla_flow_neo2b_fulllayer_libero_m323_weights.ckpt.shards.step000005000",
    "source": "ema",
    "bytes": 2529812442
  }
]