ALAM + pi0 post-trained for MetaWorld MT50
Repository ID: Mark-ZJTang/alam_plus_pi_metaworld_mt50.
Inference-only step-30,000 Orbax checkpoint for the released MetaWorld MT50
evaluation. The release configuration uses raw action horizon 6, effective
inference horizon 5, replan step 5, and camera position (0.70, 0.075, 0.70).
The public full post-training resource contract is 8 GPUs. Inference and the documented simulator acceptance use one sufficiently large idle GPU.
The checkpoint includes params, _CHECKPOINT_METADATA, and
assets/metaworld_mt50/norm_stats.json; optimizer train_state is intentionally
excluded. The matching external ALAM tokenizer is
metaworld_epoch19_step58216 from the ALAM tokenizer model repository.
Verify all files against evaluation/WEIGHTS_MANIFEST.sha256 in the GitHub code
release. License metadata must be completed before public publication.
From the matching GitHub code checkout:
.venvs/publish/bin/python workflows/publishing/download_huggingface.py \
--artifact alam_pretrain \
--artifact alam_plus_pi_metaworld_mt50
bash workflows/metaworld_evaluation/evaluate_mt50.sh --gpu 0
The frozen 50-task protocol uses 10 episodes per task. The recorded result is 434/500 episode successes (86.8% episode-weighted) and 84.98% after macro averaging Easy/Medium/Hard/Very-Hard groups, reported as 85.0%.