Better hashing for constant pool #27733
Closed
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
facebook-github-bot
added
the
oncall: jit
Contributor
facebook-github-bot
left a comment
facebook-github-bot
left a comment
There was a problem hiding this comment.
@houseroad has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
Member
|
Cool! Can you paste of a snippet of a graph that would be affected by excessive hash collisions? Also, have you tried this PR on some of the models with really long compile times? |
houseroad
force-pushed
the
fix_constant_pooling
branch
from
c36eb9c to
d8a58b3
Compare
Contributor
facebook-github-bot
left a comment
facebook-github-bot
left a comment
There was a problem hiding this comment.
@houseroad has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
houseroad
changed the title
suo
approved these changes
Oct 11, 2019
Member
suo
left a comment
suo
left a comment
There was a problem hiding this comment.
Well, overall this looks fine to me.
Contributor
|
@houseroad merged this pull request in eb52223. |
thiagocrepaldi
pushed a commit
to thiagocrepaldi/pytorch
that referenced
this pull request
Feb 4, 2020
Summary: Some models many contain thousands constants (like list of ints) and Constant Pooling and CSE pass will move the constant around and update the constant pooling. However our existing hash function only consider the node type + input type + output node (https://bddppq.github.io/codebrowser/pytorch/pytorch/torch/csrc/jit/node_hashing.cpp.html#_ZNK5torch3jit8HashNodeclEPKNS0_4NodeE), which will have a lot of conflicts... I have profiled, one insert may take as long as about 0.2 second... And loading the model will take 200 second, which is insane. So we should fix this performance issue by considering the constant value as well to avoid the conflict. Pull Request resolved: pytorch#27733 Reviewed By: bddppq Differential Revision: D17873733 Pulled By: houseroad fbshipit-source-id: 2338d7bf67174a8e56caa19a30401199f68b592a
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 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.
Some models many contain thousands constants (like list of ints) and Constant Pooling and CSE pass will move the constant around and update the constant pooling.
However our existing hash function only consider the node type + input type + output node (https://bddppq.github.io/codebrowser/pytorch/pytorch/torch/csrc/jit/node_hashing.cpp.html#_ZNK5torch3jit8HashNodeclEPKNS0_4NodeE), which will have a lot of conflicts... I have profiled, one insert may take as long as about 0.2 second... And loading the model will take 200 second, which is insane.
So we should fix this performance issue by considering the constant value as well to avoid the conflict.