feat: Fixed metric calculation and made all grpo metrics token-level #373
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…ther or Signed-off-by: Sahil Jain <sahilj@nvidia.com>
Signed-off-by: Sahil Jain <sahilj@nvidia.com>
Signed-off-by: Sahil Jain <sahilj@nvidia.com>
ashors1
reviewed
May 13, 2025
ashors1
reviewed
May 13, 2025
Signed-off-by: Sahil Jain <sahilj@nvidia.com>
Signed-off-by: Sahil Jain <sahilj@nvidia.com>
ashors1
previously approved these changes
May 14, 2025
ashors1
reviewed
May 14, 2025
parthchadha
reviewed
May 14, 2025
parthchadha
previously approved these changes
May 14, 2025
…seq counts Signed-off-by: Sahil Jain <sahilj@nvidia.com>
terrykong
previously approved these changes
May 14, 2025
Signed-off-by: Sahil Jain <sahilj@nvidia.com>
terrykong
approved these changes
May 15, 2025
YzjiaoNvd
pushed a commit
to YzjiaoNvd/NeMo-RL
that referenced
this pull request
Jun 10, 2025
…VIDIA-NeMo#373) Signed-off-by: Sahil Jain <sahilj@nvidia.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
I noticed that the loss_type of the GRPO loss was impacting how metrics were calculated. Changed all losses to always accept both global token and sequence counts so that metrics can be calculated based on whichever they should be. Also fixed a merge artifact that was causing sampling_importance_weights to be miscalculated.