pusht_96_norm4_noncot_chunk_k10_world_model_20260622_perseg (EMA, bf16)

Fine-tuned BAGEL-7B-MoT action-conditioned visual world model — PushT (96px, norm4, JPEG q90; coverage task, no hard split — in-dist claims only). Step-5000 EMA weights, cast fp32 → bfloat16 (inference runs in bf16 autocast — lossless for inference). No optimizer state; not a training-resume checkpoint.

• Load into the BAGEL-7B-MoT architecture (base: ByteDance-Seed/BAGEL-7B-MoT).
• Training data: ultrastar111/pusht_96_norm4_noncot_chunk_k10_20260622_perseg
• Recipe: cold start from vanilla BAGEL, lr 2e-5, cosine, warmup 300, 5000 steps, 40k-token packing (maze2d cot-K1: 24k), EMA 0.993.
• Eval contract: CoT ckpts need the per-segment decoder (BAGEL_DECODER=perseg) + CFG off.
• License: CC-BY-NC-4.0 (research use).

training_data_manifest.txt

STUDY:    PushT K-action-chunk feedback-interval (noncot, K=10, GT env-feedback re-grounding except singleshot)
data:     /data/home/raychai/hf_datasets/pusht_96_norm4_noncot_chunk_k10_20260622_perseg/training (q90; PushT instruction baked in by remap)
config:   ./data/configs/vlm_gym_pusht_imagined_cot_envfb_train.yaml (envfb loader 'interleaved_cot_envfb', data-dir overridden)
init:     VANILLA /home/jiaxin/unified_world_model/pretrained/BAGEL-7B-MoT (EMA); hparams lr2e-5/warmup300/cosine/total5000/tokens40000/ema0.993/save2500; OOD probe OFF
git_rev:  f7479e79b57d4a8780998f1b8282bf76fc13f906