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Outdated suggestions cannot be applied.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.
Context
What is the purpose of this PR? Is it to
addressing compile graph breaks
Test plan
Rough numbers:
11b, compiling only decoder
28s on the first step
21s on the second step
0.9s any other step
~28s in the first step,
~ 0.9s any other step
h = self.tok_embeddings(tokens)
h.requires_grad = True
First step ~18s
~ 0.9s any other step
So we save ~21s to 31s of compilation time by avoiding the graph breaks due to shape plus optionally adding requires_grad. This is without compiling the encoder, which NaNs if we do (Brian will look into that).
TODO: Need to check if there is a perf impact when packed=True and there wont be graph breaks.