twanghcmut's picture
checkpoint bundle
b05fc58 verified
|
Raw History Blame Contribute Delete
716 Bytes
metadata
license: other
tags:
  - robotics
  - vision-language-action
  - bridgequant

bridgequant-robotwin-76m

BridgeQuant 76M trained jointly on three RoboTwin 2.0 tasks (beat_block_hammer, stack_blocks_two, place_empty_cup; 50 clean demos each; relative arm joints). Multi-task joint training reached 0% in the simulator with this budget; the single-task 7M (bridgequant-robotwin-7m-hammer) is the working RoboTwin checkpoint.

Files: model.safetensors, norm_stats.json, bridgequant_vla.yaml / config.json / metadata.json (the bundle load_trainable_weights expects), train_config.yaml (the experiment YAML; evaluation and deployment use the same YAML).

Code: https://github.com/VLAOpt/bridgequant-async