sokoban_easy_cot_chunk_kinf_forward_random_preview
BAGEL-7B-MoT world-model SFT checkpoint (EMA weights only, raw fp32).
- task: sokoban_easy variant: forward_random_preview step: 4000
- Compare-two SFT with thinking randomly decorrelated from need (reinforcement-learning initialization). PREVIEW at step 4000; the step-5000 checkpoint will replace it.
- weights: fp32 EMA, sharded into 2 safetensors files (
ema-XXXXX-of-00002.safetensors)
plus ema.safetensors.index.json (fp32 exceeds the 50GB single-file limit). Optimizer state NOT
included — for inference / eval, not resume.
Load
This is an SFT delta on the public BAGEL-7B-MoT base. Load the base config / tokenizer / VAE / ViT
from BAGEL-7B-MoT together with these EMA weights; point the loader at the repo folder (or its
ema.safetensors.index.json) and it assembles the shards, e.g. with the eval runners:
--bagel-path <this repo> --bagel-base-path <BAGEL-7B-MoT>.
Image transforms (sokoban): VAE 512/256, ViT 336/224, JPEG q95, 256px render. Prompt is stop-required; CoT is all-step interleaved.
