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

[70B-Part1] Prefetch weights separately #106

Merged
merged 16 commits into from
Sep 13, 2024
Merged

[70B-Part1] Prefetch weights separately #106

merged 16 commits into from
Sep 13, 2024

Conversation

farzadab
Copy link
Contributor

This PR allows for downloading the weights of the model first, before starting the DDP processes. The previous approach tends to time out when using 70B model.

@farzadab
Copy link
Contributor Author

farzadab commented Sep 10, 2024

@liPatrick I'm gonna add you to the PRs for visibility. You don't have to do the review.

mcloud.yaml Outdated Show resolved Hide resolved
Copy link
Contributor

@juberti juberti left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LG, just nits.

ultravox/training/helpers/prefetch_weights.py Outdated Show resolved Hide resolved
ultravox/training/helpers/prefetch_weights.py Outdated Show resolved Hide resolved
ultravox/training/train.py Outdated Show resolved Hide resolved
@farzadab farzadab merged commit 7bd4eeb into main Sep 13, 2024
1 check passed
@farzadab farzadab deleted the farzad-fsdp-p1 branch September 13, 2024 22:40
akshat0311 pushed a commit to jiviai/audio-llm that referenced this pull request Jan 30, 2025
* prefetch weights separately

* moving get_train_args to config

* add a test for prefetch_weights

* double check for prefetch_weights for local/test runs
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants