M. Ernestus

Upvote for this PR. It is tiny and it fixes a large issue which prevents many users from upgrading to 3.5

I have the issue too: **Context:** Global generic exception ```python-traceback Traceback (most recent call last): File "/app/lib/python3.9/site-packages/GTG/gtk/browser/main_window.py", line 1060, in on_edit_active_task self.app.open_task(tid) File "/app/lib/python3.9/site-packages/GTG/gtk/application.py", line 553, in open_task editor =...

I have the same issue. Additionally, my computer freezes for about half a minute when I open the help page :grimacing:

I did this: ```zsh ~ ➤ /usr/bin/flatpak run --branch=stab~ ➤ /usr/bin/flatpak run --branch=sta~ ➤ /usr/bin/flatpak run --branch=stable --arch=x86_64 --command=gtg --file-forwarding org.gnome.GTG Gtk-Message: 10:47:15.113: Failed to load module "xapp-gtk3-module" ``` Note:...

I am a bit stuck right now and hope to get some input. Maybe @AdamGleave or @araffin can help? The first issue is with the helpers in `imitation.policies.serialize`. They impose...

> What command can I run to replicate this failure? In CircleCI logs I see some scripts failing, but nothing that seemed preference comparison specific. Seems like the CI tests...

I see it makes sense to integrate the huggingface loading logic into `serialize.py`, especially if it is ok to make breaking changes. When loading from huggingface we pull the model...

Concerning demonstrations: I agree we still want to load demonstrations from disk. Right now there is no canonical way to upload RL-demonstrations to huggingface hub but there is the notion...

Right now the remaining failing tests are `test_scripts.py::test_analyze_imitation` and `test_scripts.py::test_convert_trajs_in_place`. Before I put time into understanding and debugging those I wanted to verify, that they still used? `test_convert_trajs_in_place` seems to...

I can easily make the `test_convert_trajs_in_place` case pass by adding back the demonstrations (`final.pkl`) in the testdata folder. However, I figured out that the `final.pkl` in the current master branch...