Reuse existing output node when replacing operator with a fusion #401
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
robertknight
force-pushed
the
optimize-reuse-output
branch
from
b004dd9
to
04559ca
Compare
When replacing a subgraph with a fused operator, reuse the output value node from the subgraph instead of creating a new one. This preserves metadata such as the name and shape associated with that value node. Also it simplifies the code by removing the need to replace all references to the previous output node with the new one. This improves the runtime of the Whisper example using the whisper-base model by ~25% by fixing an issue where fused Transpose + MatMul operations did not get used.
Exposing `add_node` as a public method creates a hazard as it allows callers to add nodes to the graph without post-processing steps that other methods do (eg. `add_op`). Make this method private again and add a more specific `add_constant_node` alternative to handle the single use case for it outside the graph module.
robertknight
force-pushed
the
optimize-reuse-output
branch
from
04559ca
to
4937d83
Compare
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Outdated suggestions cannot be applied.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.
When replacing a subgraph with a fused operator, reuse the output value node
from the subgraph instead of creating a new one. This preserves metadata such as
the name and shape associated with that value node. Also it simplifies the code
by removing the need to replace all references to the previous output node with
the new one.
This improves the runtime of the Whisper example using the whisper-base model by
~25% by fixing an issue where fused Transpose + MatMul operations did not get
used.