Improve NEST GPU Utilization 2/N #14089
Merged
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
stevehuang52
reviewed
Jul 1, 2025
| mask_value = self.mask_embedding.unsqueeze(-1) | ||
| masks = torch.zeros_like(input_feats) | ||
| maksed_feats = input_feats.clone() | ||
| masked_feats = input_feats |
There was a problem hiding this comment.
Will this also change input_feats as well when masked_feats is updated? Ideally we should return the masked_feats while keeping input_feats unchanged.
There was a problem hiding this comment.
I did confirm that by:
module = RandomBlockMasking(80, allow_overlap=False, mask_prob=0.01, block_size=40).cuda()
for i in range(100):
input = torch.randn(1, 80, 1000).cuda()
input_len = torch.tensor([1000]).cuda()
masked_feats, masks = module(input, input_len)
assert not torch.allclose(masked_feats, input)
|
@MahmoudAshraf97 could you please fix the DCO error? |
MahmoudAshraf97
force-pushed
the
nest_masking
branch
from
5559d1a to
2e5e1b9
Compare
Signed-off-by: MahmoudAshraf97 <hassouna97.ma@gmail.com>
MahmoudAshraf97
force-pushed
the
nest_masking
branch
from
2e5e1b9 to
545b55f
Compare
|
@stevehuang52 Fixed |
|
@stevehuang52 Can you rerun CICD? |
stevehuang52
approved these changes
Jul 7, 2025
AmirHussein96
pushed a commit
to AmirHussein96/NeMo
that referenced
this pull request
Jul 23, 2025
Signed-off-by: MahmoudAshraf97 <hassouna97.ma@gmail.com> Signed-off-by: Amir Hussein <amhussein@nvidia.com>
AmirHussein96
pushed a commit
to AmirHussein96/NeMo
that referenced
this pull request
Aug 5, 2025
Signed-off-by: MahmoudAshraf97 <hassouna97.ma@gmail.com> Signed-off-by: Amir Hussein <amhussein@nvidia.com>
AmirHussein96
pushed a commit
to AmirHussein96/NeMo
that referenced
this pull request
Aug 5, 2025
Signed-off-by: MahmoudAshraf97 <hassouna97.ma@gmail.com> Signed-off-by: Amir Hussein <amhussein@nvidia.com>
nasretdinovr
pushed a commit
to nasretdinovr/NeMo
that referenced
this pull request
Aug 8, 2025
Signed-off-by: MahmoudAshraf97 <hassouna97.ma@gmail.com>
guyueh1
pushed a commit
to guyueh1/NeMo
that referenced
this pull request
Aug 25, 2025
Signed-off-by: MahmoudAshraf97 <hassouna97.ma@gmail.com> Signed-off-by: Guyue Huang <guyueh@nvidia.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 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.
This PR is among a series aimed to improve the training speed and GPU utilization of NEST models #13619
after profiling, the masking module takes a good chunk of the total training loop time while involving no computation, thus reducing the utilization, this PR reduces python for-loops as much as possible, and converts the masking operation to a singe function call
As shown in the following graph, This PR along with #14086 reduces the training step time by almost 17%
The green is with the PRs applied
cc @stevehuang52