Use latest apex internal API #8129
Merged
Conversation
This file contains 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
mikolajblaz
approved these changes
Jan 9, 2024
jbaczek
changed the title
8 tasks
Signed-off-by: Jan Baczek <jbaczek@nvidia.com>
janekl
force-pushed
the
jbaczek/bugfix/update_LayerNorm1P_API
branch
from
143aab7 to
4d18d26
Compare
| @@ -40,7 +40,7 @@ def reset_parameters(self): | |||
| torch.nn.init.zeros_(self.bias) | |||
|
|
|||
| def forward(self, x): | |||
| return _fast_layer_norm(x, self.weight + 1, self.bias, self.epsilon) | |||
| return _fast_layer_norm(x, self.weight + 1, self.bias, self.epsilon, memory_efficient=False) | |||
There was a problem hiding this comment.
why is this hardcoded to False? Instead it can be run by just passing self.memory_efficient as a parameter right?
There was a problem hiding this comment.
AFAIU memory_efficient=False is the previous implementation of _fast_layer_norm. As non memory efficient implementation has been tested I left it like this.
We can make it switchable but I think it would be beneficial to test it first. What do you think?
There was a problem hiding this comment.
Sounds good, we can keep it as it is
ShriyaPalsamudram
approved these changes
Jan 16, 2024
layalir
approved these changes
Jan 16, 2024
minitu
pushed a commit
to minitu/NeMo
that referenced
this pull request
Jan 19, 2024
Signed-off-by: Jan Baczek <jbaczek@nvidia.com>
ssh-meister
pushed a commit
to ssh-meister/NeMo
that referenced
this pull request
Feb 15, 2024
Signed-off-by: Jan Baczek <jbaczek@nvidia.com> Signed-off-by: Sasha Meister <ameister@nvidia.com>
pablo-garay
pushed a commit
that referenced
this pull request
Mar 19, 2024
Signed-off-by: Jan Baczek <jbaczek@nvidia.com> Signed-off-by: Pablo Garay <pagaray@nvidia.com>
rohitrango
pushed a commit
to rohitrango/NeMo
that referenced
this pull request
Jun 25, 2024
Signed-off-by: Jan Baczek <jbaczek@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.
What does this PR do ?
Apex _fast_layer_norm added mandatory argument
memory_efficient. This PR fixes call to this function to accommodate to the change.Jenkins CI
To run Jenkins, a NeMo User with write access must comment
jenkinson the PR.Before your PR is "Ready for review"
Pre checks:
PR Type:
Who can review?
@mikolajblaz can you review this one? I can see that you introduced this line 17 months ago. Can we move to the APEX implementation entirely as suggested in the TODO?
Additional Information