Archive: calib-phase1
EMA-consolidated research checkpoint archive.
NEO in-context scale-calibration Phase-1 finetunes (base + cam00/01/02/05/10). Verdict: crop-zoom aug alone fixes zoom (95.2@z1.5); oracle calibration token HURTS (-7.4, monotone in z).
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_calib_base_calib_base",
"step_dir": "lafm_calib_base_weights.ckpt.shards.step000005000",
"source": "ema",
"bytes": 1774687498
},
{
"run": "lafm_calib_cam00_calib_cam00",
"step_dir": "lafm_calib_cam00_weights.ckpt.shards.step000005000",
"source": "ema",
"bytes": 1774687498
},
{
"run": "lafm_calib_cam01_calib_cam01",
"step_dir": "lafm_calib_cam01_weights.ckpt.shards.step000005000",
"source": "ema",
"bytes": 1774687498
},
{
"run": "lafm_calib_cam02_calib_cam02",
"step_dir": "lafm_calib_cam02_weights.ckpt.shards.step000005000",
"source": "ema",
"bytes": 1774687498
},
{
"run": "lafm_calib_cam05_calib_cam05",
"step_dir": "lafm_calib_cam05_weights.ckpt.shards.step000005000",
"source": "ema",
"bytes": 1774687498
},
{
"run": "lafm_calib_cam10_calib_cam10",
"step_dir": "lafm_calib_cam10_weights.ckpt.shards.step000005000",
"source": "ema",
"bytes": 1774687498
}
]
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support