-
Notifications
You must be signed in to change notification settings - Fork 413
[Example] Multiagent examples: MAPPO-IPPO-MADDPG-IDDPG-IQL-QMIX-VDN #1027
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
Signed-off-by: Matteo Bettini <[email protected]>
I have addressed the comments in the examples, when you can have a look. I moved the models and loss to #1378 so we can merge them first |
Signed-off-by: Matteo Bettini <[email protected]>
Signed-off-by: Matteo Bettini <[email protected]>
Signed-off-by: Matteo Bettini <[email protected]>
Signed-off-by: Matteo Bettini <[email protected]>
# Conflicts: # torchrl/objectives/__init__.py # torchrl/objectives/multiagent/qmixer.py
e3fbc2e
to
dd0732a
Compare
Signed-off-by: Matteo Bettini <[email protected]>
Signed-off-by: Matteo Bettini <[email protected]>
Signed-off-by: Matteo Bettini <[email protected]>
Signed-off-by: Matteo Bettini <[email protected]>
# Conflicts: # test/test_cost.py # torchrl/modules/models/multiagent.py # torchrl/objectives/dqn.py # torchrl/objectives/multiagent/qmixer.py
Signed-off-by: Matteo Bettini <[email protected]>
examples/media/marl_vmas.png
Outdated
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I would recommend not uploading media to this repo.
The repo size is already huge and this can impact user with limited disk space.
I would recommend creating a secundary repo/s3 bucket/ some other hosting service to host the media and just embed links to them
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I agree, this isn't scalable on the long term.
We need to make sure that whatever we link will remain on the long term
pytorch/tutorials and pytorch/pytorch (as well as vision I think) are pretty big though
Let me see how we're supposed to handle things
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
yeah i dunno if pytorch has any hosting service or repo, but those would be cool.
It is your call anyway, I was just pointing out that git will not only save the media, but also all their previous versions, and it will all be downloaded upon clone
Signed-off-by: Matteo Bettini <[email protected]>
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
we must add these examples to the example CI
They should not take more than a minute each to run (ie very small nets and very few iterations). We just want to make sure they run ok
I'll what we can do regarding the images and s3 bucket
examples/media/marl_vmas.png
Outdated
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I agree, since this file is likely to be modified in the future, to not make the git history too big we should host it on our s3 bucket
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Fine by me :)
Signed-off-by: Matteo Bettini <[email protected]>
The examples are now in the CI, if we sort out the s3 we are ready to go |
Signed-off-by: Matteo Bettini <[email protected]>
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM
The figure can be found here
Signed-off-by: Matteo Bettini <[email protected]>
updated |
…ytorch#1027) Signed-off-by: Matteo Bettini <[email protected]>
Example of multiagent training algos in vmas