Hi @RGring , very interesting work! To allow the community to use your pre-trained models, it can be helpful to be able to load them without ROS, and using the main stable-baselines distribution. This PR attempts to do just that, and being small, it should not introduce any bugs.
The one thing I'm worried about, is that the conv1d layer function added to common_custom_policies module overshadows the declaration in your stable-baselines fork. It should not cause any issues, but could cause confusion in the future. I let you decide if/how you'd like to handle this.
Hi @RGring , very interesting work! To allow the community to use your pre-trained models, it can be helpful to be able to load them without ROS, and using the main stable-baselines distribution. This PR attempts to do just that, and being small, it should not introduce any bugs.
The one thing I'm worried about, is that the conv1d layer function added to common_custom_policies module overshadows the declaration in your stable-baselines fork. It should not cause any issues, but could cause confusion in the future. I let you decide if/how you'd like to handle this.
Feel free to reject this PR if it does not LGTY!