Motphys/Uni-SuperMimic

选用指南

AI 根据 README 整理 · 2026/10/23 13:56:21

使用方式

通过命令行使用 uv 运行训练、评估和视频导出脚本，例如 `unisupermimic-train`。

查看 README 依据

│ │ └── registry.py # G1SuperMimic registration │ ├── scripts/train_supermimic.py # Training, resume, and evaluation entry point │ ├── conf/supermimic/ # Hydra configuration │ ├── assets/robots/g1/ # G1 MJCF; meshes/URDF are downloaded and untracked │ ├── assets/motions/ # LAFAN1 CSV files, downloaded and untracked │ └── sim2sim.py # Policy I/O, network shape, and asset identity checks ├── docs/assets/ # Result curves, MP4, and GIF ├── tools/prepare_supermimic_assets.py ├── tests/

uv run tools/prepare_supermimic_assets.py --validate-only ``` Assets come from `lvhaidong/LAFAN1_Retargeting_Dataset`. A complete asset set contains 40 CSV files with 264,705 source frames and 36 columns per frame. Meshes, URDF, CSV files, logs, and checkpoints are excluded by `.gitignore`. ## Training Run the following commands from the repository root and set `CUDA_VISIBLE_DEVICES` to an available GPU when using CUDA. ```bash # Default: mjwarp, 4096 parallel environments, 12000 iterations. CUDA_VISIBLE_DEVICES=0 uv run unisupermimic-train \ training.log_dir=runs/supermimic_mjwarp_001

部署要求

需要 Linux、Python 3.10–3.13 和 uv；训练 mjwarp 后端需要 CUDA GPU，AMD ROCm 主机可用预解析的 ROCm lockfile。

查看 README 依据

├── pyproject.toml, uv.lock ├── LICENSE, LICENSES/, NOTICE └── Makefile ``` ## Installation Requirements: Linux, Python 3.10–3.13, and [uv](https://docs.astral.sh/uv/). A CUDA GPU is required for mjwarp training. Run all commands from the repository root. `uv run` uses `uv.lock` to synchronize the project, PyPI `unilab`, all three simulation backends, and development dependencies. ```bash uv run unisupermimic-train --help ``` Run `uv lock` after changing dependencies. On AMD ROCm hosts, install from the pre-resolved ROCm lockfile instead (a ROCm

PyTorch build with the same pinned dependency set): ```bash make sync-rocm ``` CUDA-only backends (mjwarp) stay unsupported there; train with the MuJoCo host-bridge backend (`task=g1_super_mimic/mujoco`). `make sync-rocm` swaps `pyproject.toml`/`uv.lock` for the ROCm profile; run `make sync-default` to switch back. Maintainers regenerate the ROCm lockfile after dependency changes with `make lock-rocm`. ### Prepare Assets ```bash # Download LAFAN1/G1 data from Hugging Face, then generate and validate MJCF. uv run tools/prepare_supermimic_assets.py # Validate existing assets offline.

已知限制

CUDA-only 后端（mjwarp）在 AMD ROCm 主机上不支持，需使用 MuJoCo 后端。

查看 README 依据

PyTorch build with the same pinned dependency set): ```bash make sync-rocm ``` CUDA-only backends (mjwarp) stay unsupported there; train with the MuJoCo host-bridge backend (`task=g1_super_mimic/mujoco`). `make sync-rocm` swaps `pyproject.toml`/`uv.lock` for the ROCm profile; run `make sync-default` to switch back. Maintainers regenerate the ROCm lockfile after dependency changes with `make lock-rocm`. ### Prepare Assets ```bash # Download LAFAN1/G1 data from Hugging Face, then generate and validate MJCF. uv run tools/prepare_supermimic_assets.py # Validate existing assets offline.