Add a functional test of frame stacking#15424
Merged
blisc merged 11 commits intoNVIDIA-NeMo:mainfrom Feb 26, 2026
Merged
Conversation
Signed-off-by: Fejgin, Roy <rfejgin@nvidia.com>
rfejgin
changed the title
blisc
requested changes
Feb 24, 2026
tests/functional_tests/L2_TTS_InferEvaluate_Magpietts_FrameStacking.sh
Outdated
Show resolved
Hide resolved
blisc
reviewed
Feb 24, 2026
Signed-off-by: Fejgin, Roy <rfejgin@nvidia.com>
…Mo into magpietts_test_frame_stacking
Signed-off-by: Fejgin, Roy <rfejgin@nvidia.com>
(debugging a test failure) Signed-off-by: Fejgin, Roy <rfejgin@nvidia.com>
Collaborator
Author
|
Attention prior parameters need to be tuned for it to work well with 4x frame stacking. Disabling it for now for this test. |
1. Enable attention prior in frame stacking test. Did some basic tuning to get the attention prior to be functional with this checkpoints. 2. Decrease the SSIM target for the longform MOE test since it was sporadically failing wiht SSIM just below the threshold. Signed-off-by: Fejgin, Roy <rfejgin@nvidia.com>
XuesongYang
reviewed
Feb 26, 2026
tests/functional_tests/L2_TTS_InferEvaluatelongform_Magpietts_MoE_ZeroShot.sh
Show resolved
Hide resolved
XuesongYang
previously approved these changes
Feb 26, 2026
It would need to be tuned further to work with 4x stacking. Signed-off-by: Fejgin, Roy <rfejgin@nvidia.com>
blisc
approved these changes
Feb 26, 2026
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Adding a functional test that verifies that frame stacked models are meeting performance expectations.
In the test, we use a model with a
frame_stacking_factorof4since that tends to uncover more corner cases than lower frame stacking factors. Local transformer inference is enabled, as required for frame-stacked models.A corresponding checkpoint has been uploaded to CI storage on Azure.
Also: loosened the SSIM threshold for the MoE Longform test since it was sporadically failing with an SSIM slightly below the target.