lerobot / hy_vla_umi

Hy-Embodied-0.5-VLA — LeRobot conversion

Byte-identical tensor conversion of tencent/Hy-Embodied-0.5-VLA-UMI at revision 3f53d1f8d2bc587c523cfdc9f1041ceee42c2524 .

• Author runtime revision: 8ba4c8cbdf42a4bcf0a19be4bd2841405dfe15e9
• Model token horizon: 50
• Physical action horizon: 50
• Actions executed per sample: 50
• Image history: 1 frames at interval 1
• Action representation: relative
• Euler inference steps: 10
• Padded state/action width: 32/32
• Tensor mapping: identity (1628 tensors)

The runtime consumes an already-selected LeRobot task string and never samples task augmentation text. norm_stats.pkl is retained for provenance; normalization is loaded at runtime from safe serialized processor state.

Validation

• Strict load, save_pretrained , and strict reload passed with all 1,628 tensors mapped and no missing, unexpected, duplicate, or shape-mismatched tensors.
• Full author-vs-LeRobot forward parity passed for batch sizes 1 and 2. Images, masks, tokens, state/targets, training velocity/loss, every velocity from the 10-step Euler sampler, the complete 11-state trajectory, and final sampled actions all had max, p95, and p99 error 0.0 .
• One official BF16 forward/backward/SGD step had finite loss and gradients, 372,388,880 nonzero gradient elements, and a nonzero parameter update.

Limitation

This policy supports the released 16D dual-arm pose/gripper embodiment only. A 12D mobile-manipulator action contains base semantics that the released 20D dual-arm representation cannot express and is rejected rather than silently truncated or reinterpreted.