dp / code /hb_baselines_v1 /baselines /__init__.py
BIue-Jay's picture
Add B6 DP batch-512 step-4000 checkpoint with resume state (2026-11-17)
ab621ab verified
Raw History Blame Contribute Delete
669 Bytes
"""FM-DiT, DP (Diffusion Policy, DDPM-DiT) and ACT visuomotor baselines of the Humanoid Benchmark.
Ported from DextraVerse's imitation-learning core (`dextraverse/algorithms/il/models/{fm_dit,ddpm_dit,act}.py` and
their networks), torch only. `train.py` trains one on a task's LeRobot v2.1 expert dataset through a robot contract
(`data.CONTRACTS`: G1 state44/action76 or H2 state78/action75); `runtime.BaselinePolicy` serves a checkpoint to the
G1 evaluation adapters (`evaluation/policy/{FMDiT,DP,ACT}/model.py`, via `adapters.py`) and to the H2 adapter
(`debug/newton/h2_policy_adapters/baseline.py`). Usage and the environment: evaluation/README.md#baselines.
"""