Enable PyTorch Models to Share Weights #4123
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
tanmayv25
requested changes
Mar 29, 2022
tanmayv25
reviewed
Mar 29, 2022
CoderHam
previously approved these changes
Mar 30, 2022
dyastremsky
force-pushed
the
dyas-share-weights
branch
from
328c199
to
d3f98ee
Compare
tanmayv25
approved these changes
Mar 30, 2022
CoderHam
approved these changes
Mar 30, 2022
|
Does weight sharing mean that multiple instances of a model will not truly run in parallel in their separate CUDA streams? |
|
@joaopcm1996 It should not have an impact. Since the models are already trained, we're just sharing the constant (read-only) values. |
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Outdated suggestions cannot be applied.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.
This PR creates tests for an associated PyTorch backend change, which allows a Triton user to enable multiple instances of a model on the same device to share weights. This is turned off by default and can be enabled via a model config parameter, "ENABLE_WEIGHT_SHARING."
Enabling weight reuse can reduce memory usage of model loading and inference. It should not be used with models that maintain state (due to the reusing of weights).
Related backend code and documentation change: triton-inference-server/pytorch_backend#54