From cc7234ad7ff39a8f32de3af903606723a16f0648 Mon Sep 17 00:00:00 2001 From: allenzren Date: Tue, 4 Feb 2025 11:50:53 -0500 Subject: [PATCH] add note about `ft_denoising_steps` in eval in README --- README.md | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/README.md b/README.md index 4c83167..cafe680 100644 --- a/README.md +++ b/README.md @@ -138,14 +138,14 @@ See [here](cfg/finetuning.md) for details of the experiments in the paper. * Videos of trials in Robomimic tasks can be recorded by specifying `env.save_video=True`, `train.render.freq=`, and `train.render.num=` in fine-tuning configs. ## Usage - Evaluation -Pre-trained or fine-tuned policies can be evaluated without running the fine-tuning script now. Some example configs are provided under `cfg/{gym/robomimic/furniture}/eval}` including ones below. `ft_denoising_steps` needs to match fine-tuning config. Set `base_policy_path` to override the default checkpoint. +Pre-trained or fine-tuned policies can be evaluated without running the fine-tuning script now. Some example configs are provided under `cfg/{gym/robomimic/furniture}/eval}` including ones below. Set `base_policy_path` to override the default checkpoint, and `ft_denoising_steps` needs to match fine-tuning config (otherwise assumes `ft_denoising_steps=0`, which means evaluating the pre-trained policy). ```console python script/run.py --config-name=eval_diffusion_mlp \ - --config-dir=cfg/gym/eval/hopper-v2 + --config-dir=cfg/gym/eval/hopper-v2 ft_denoising_steps=? python script/run.py --config-name=eval_{diffusion/gaussian}_mlp_{?img} \ - --config-dir=cfg/robomimic/eval/can + --config-dir=cfg/robomimic/eval/can ft_denoising_steps=? python script/run.py --config-name=eval_diffusion_mlp \ - --config-dir=cfg/furniture/eval/one_leg_low + --config-dir=cfg/furniture/eval/one_leg_low ft_denoising_steps=? ``` ## DPPO implementation