Skip to content
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

Config instantiation is not DDP-aware #83

Closed
surajpaib opened this issue Oct 10, 2023 · 5 comments
Closed

Config instantiation is not DDP-aware #83

surajpaib opened this issue Oct 10, 2023 · 5 comments
Labels
bug Something isn't working

Comments

@surajpaib
Copy link
Collaborator

Because config is instantiated only on the main process and then DDP is initialized, objects that are initialized in the config are not DDP aware.

Maybe there is a way to fix this by instantiating trainer first and separate and then the system

@surajpaib surajpaib added the bug Something isn't working label Oct 10, 2023
@ibro45
Copy link
Collaborator

ibro45 commented Jan 16, 2024

I think this isn't an issue anymore, right? Closing.

@ibro45 ibro45 closed this as completed Jan 16, 2024
@surajpaib surajpaib reopened this Apr 17, 2024
@surajpaib
Copy link
Collaborator Author

@ibro45 This is still an issue.

Now I needed this for Lightning-AI/pytorch-lightning#11763

But since at instantiation, objects are not DDP aware, this problem persists.

@ibro45
Copy link
Collaborator

ibro45 commented Apr 18, 2024

Any idea on how to approach it?

@ibro45
Copy link
Collaborator

ibro45 commented Jun 15, 2024

Wait how is this a lighter problem? I think it has to do with improper usage/behavior of CacheDataset only

@ibro45
Copy link
Collaborator

ibro45 commented Jun 15, 2024

Would swapping the order of trainer and system help?

trainer = parser.get_parsed_content("trainer")

@ibro45 ibro45 closed this as completed Jan 9, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
bug Something isn't working
Projects
None yet
Development

No branches or pull requests

2 participants