NVIDIA / NeMo

A scalable generative AI framework built for researchers and developers working on Large Language Models, Multimodal, and Speech AI (Automatic Speech Recognition and Text-to-Speech)
https://docs.nvidia.com/nemo-framework/user-guide/latest/overview.html
Apache License 2.0
12.14k stars 2.53k forks source link

fix: cosine annealing linear warmup lr #10530

Closed clumsy closed 1 month ago

clumsy commented 2 months ago

What does this PR do ?

CosineAnnealing LR with linear warmup and constant_steps is 0 on the first step. This causes issues for AdamW optimizer which uses LR for step_size in the denominator. This is due to the fact on first step when this method is called step=0 before the first optimizer step is complete.

Using the same protection as in:

Updated unit tests not to tolerate LR=0.

Collection: [Note which collection this PR will affect]

Core

Changelog

Usage

# Add a code snippet demonstrating how to use this 

GitHub Actions CI

The Jenkins CI system has been replaced by GitHub Actions self-hosted runners.

The GitHub Actions CI will run automatically when the "Run CICD" label is added to the PR. To re-run CI remove and add the label again. To run CI on an untrusted fork, a NeMo user with write access must first click "Approve and run".

Before your PR is "Ready for review"

Pre checks:

PR Type:

If you haven't finished some of the above items you can still open "Draft" PR.

Who can review?

Anyone in the community is free to review the PR once the checks have passed. Contributor guidelines contains specific people who can review PRs to various areas.

Additional Information