Download eval_scene.py from lzhts1/ProWorld: direct link, hf CLI and curl.
- Browser
- Download file 639 Bytes
-
https://huggingface.co/datasets/lzhts1/ProWorld/resolve/main/eval_scene.py
- Command line
-
hf download hf://datasets/lzhts1/ProWorld/eval_scene.py
-
curl -L -o eval_scene.py https://huggingface.co/datasets/lzhts1/ProWorld/resolve/main/eval_scene.py
639 Bytes
| """ | |
| Evaluate the original LeWM model on OGBench Scene. | |
| Example: | |
| MUJOCO_GL=egl python eval_scene.py \ | |
| policy=/data_nvme/user/zliu681/le-wm-main/lewm_cache/ogbench_scene/lewm_scene/lewm_scene_epoch_10 \ | |
| solver.device=auto \ | |
| eval.num_eval=50 \ | |
| eval.save_video=True | |
| """ | |
| import faulthandler | |
| import sys | |
| from eval import run | |
| def _has_config_name(argv: list[str]) -> bool: | |
| return any(arg == "--config-name" or arg.startswith("--config-name=") for arg in argv) | |
| if __name__ == "__main__": | |
| faulthandler.enable(all_threads=True) | |
| if not _has_config_name(sys.argv): | |
| sys.argv.insert(1, "--config-name=scene") | |
| run() | |