Thomas Lips
Thomas Lips
When using `share_rgb_model` to share the weights across different image inputs, the Batchnorms are not replaced by groupnorms in the [`MultiImageObsEncoder` ](https://github.com/real-stanford/diffusion_policy/blob/5ba07ac6661db573af695b419a7947ecb704690f/diffusion_policy/model/vision/multi_image_obs_encoder.py#L38). This results in very unstable training and reduced...
Hi, thanks for building this benchmark. I am looking to add some object shape variations (not only scale). Any insights on the most elegant way to incorporate this in your...
Hi everyone Does anyone know of a similar project for imitation learning algorithms? (think diffusion policy, ACT, ...). Lerobot combines a number of implementations, but does not have the same...