ppo-LunarLander-v2 model from unit 8.1 overwriting ppo-LunarLander-v2 model from unit 1 in results
Hi Thomas,
(Having written a question then resolved it myself)
I'm just writing to document the way I discovered to handle models overwriting one another when the repo name is identical.
Basically, prevention is better than cure, use different repo names for unit 1 & unit 8.1 is the easiest;
However, to 'rescue' an overwrite, these are the basic stages
1 Use git or the huggingface api to download the unit 1 commit to a local git repository;
1A (optionally?) Move that commit to a new named branch locally
2 Create a new huggingface repository with a different name;
3 Set the new remote repository to track the (new named) local branch
4 Push the local branch to the new remote repository
Having looked at the app.py code for Check-my-progress-Deep-RL-Course clearly the keys that identify an individual model are not its repo name, rather its env name together with its lib name; that suggested simply putting the unit 1 commit into a different repo so that the tip of the new repo would return the unit 1 env/lib key combination, while the tip of the original repo returns the unit 8.1 env/lib combination
Thanks.
Faced the same issue today.
Your guide was very helpful!