[jit] Fix scalar tensor assert in fusion compiler#10952
Closed
zou3519 wants to merge 6 commits intopytorch:masterfrom
Closed
[jit] Fix scalar tensor assert in fusion compiler#10952zou3519 wants to merge 6 commits intopytorch:masterfrom
zou3519 wants to merge 6 commits intopytorch:masterfrom
Conversation
Contributor
facebook-github-bot
left a comment
There was a problem hiding this comment.
zou3519 has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
Contributor
|
@zou3519 : This fixes the particular |
Contributor
Author
|
Oh I see, yes I think that is related. Sending a fix... |
Contributor
|
Thanks! I think that this fixes the scalar issue, and the Hamiltonian Monte Carlo examples pass with this change. The VAE example is now throwing a different error, but I think that might be a separate issue. I'll either create a new one or update #10715. Stack Trace |
Contributor
Author
|
@neerajprad the stack track you have there is probably a different issue. Let me know (or update/open an issue) if you can find a small example that causes the crash |
Contributor
Thanks, @zou3519! I'll debug to see if there is a minimal example that we can isolate. |
apaszke
reviewed
Aug 31, 2018
torch/csrc/jit/fusion_compiler.h
Outdated
This comment was marked as off-topic.
This comment was marked as off-topic.
Sorry, something went wrong.
torch/csrc/jit/fusion_compiler.cpp
Outdated
This comment was marked as off-topic.
This comment was marked as off-topic.
Sorry, something went wrong.
This comment was marked as off-topic.
This comment was marked as off-topic.
Sorry, something went wrong.
This comment was marked as off-topic.
This comment was marked as off-topic.
Sorry, something went wrong.
Contributor
Author
|
I can sit on this until @mruberry's fuser split happens :) |
Contributor
|
@zou3519 : Regarding my comment above, I think that is most likely a Pyro specific issue pyro-ppl/pyro#1358. |
Collaborator
|
I can rebase my PR with your changes. My PR is still waiting on review and I don't want to hold things up. |
Contributor
Author
|
Thanks, @mruberry, I'll let you know when this goes in. @neerajprad based on the error message it looks like the tracer is specializing on input sizes -- I think we changed the behavior last week and it shouldn't do that anymore. Could you try pulling from master again and see if your bug still exists? |
Contributor
I have verified that the PyTorch tracer correctly generalizes to other input sizes (at least for the examples I tried it on), so this is most likely a bug within Pyro (in the way we are using the tracing functionality). This is being tracked in pyro-ppl/pyro#1358. |
apaszke
reviewed
Sep 4, 2018
torch/csrc/jit/fusion_compiler.cpp
Outdated
This comment was marked as off-topic.
This comment was marked as off-topic.
Sorry, something went wrong.
Contributor
facebook-github-bot
left a comment
There was a problem hiding this comment.
zou3519 has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
apaszke
approved these changes
Sep 4, 2018
Contributor
|
My only concern is that this doesn't really check that the fusion even happened on those scalars, but LGTM. |
Contributor
facebook-github-bot
left a comment
There was a problem hiding this comment.
zou3519 has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
Fixes pytorch#8560. Unblocks pytorch#10715. The assert (nDim <= uncompressedDims) was being triggered for a scalar tensor because we compute nDim to be 1 for a scalar tensor but uncompressedDim = 0. This PR changes it so that we compute nDim to be 0 for a scalar tensor. This works because indexing in a kernel depends on nDim. If nDim = 0, then offset is always 0, which is what we want. Some other (small) changes were necessary to make this work: - One cannot define a 0-length array `IndexType arr[0]` so the code guards against that - Needed to change some of the maxTensorInfoSize logic to handle the case when uncompressedDim == 0.
0899410 to
78651c2
Compare
Contributor
facebook-github-bot
left a comment
There was a problem hiding this comment.
zou3519 has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
Contributor
facebook-github-bot
left a comment
There was a problem hiding this comment.
zou3519 has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
petrex
pushed a commit
to petrex/pytorch
that referenced
this pull request
Sep 6, 2018
* upstream/master: (26 commits) cudnn 7 upgrade with spatialBN fix (pytorch#11291) Ignore FuseGraph Call on Windows (pytorch#11015) defer resolution of mkl to a cmake wrapper library (pytorch#11298) Cleanup dependency of distributed flags (pytorch#11221) Move minimal wrapdim functionality to core, remove THTensor include i… (pytorch#11283) Change includes from ATen/Storage.h to ATen/core/Storage.h (pytorch#11217) Fix scalar tensor assert in fusion compiler (pytorch#10952) Add dead code elimination pass (pytorch#10101) Distributed Data Parallel CPU module for C10D (pytorch#11168) Back out "[pt1][tensor] Add strides to caffe2::Tensor" Fix conv gradient conversion (pytorch#11312) Bag of clang tidy fixes for torch/csrc/ and torch/csrc/autograd (pytorch#11050) Sparse tensor printing; add NotImplemented autograd fn (pytorch#10181) Add convertToCaffe2Proto to python API fix doc for functional.dropout* (pytorch#10417) typo fix Tranpose2D -> Transpose2D (pytorch#11281) Remove THFinalizer Forward declarations of needed curand functions (pytorch#10911) nomnigraph - simplify core graph API and test (pytorch#11256) Small fixes to cppdocs for sync script (pytorch#11300) ...
PenghuiCheng
pushed a commit
to PenghuiCheng/pytorch
that referenced
this pull request
Sep 11, 2018
Summary: Fixes pytorch#8560. Unblocks pytorch#10715. The assert (nDim <= uncompressedDims) was being triggered for a scalar tensor because we compute nDim to be 1 for a scalar tensor but uncompressedDim = 0. This PR changes it so that we compute nDim to be 0 for a scalar tensor. This works because indexing in a kernel depends on nDim. If nDim = 0, then offset is always 0, which is what we want. Some other (small) changes were necessary to make this work: - One cannot define a 0-length array `IndexType arr[0]` so the code guards against that - Needed to change some of the maxTensorInfoSize logic to handle the case when uncompressedDim == 0. cc apaszke zdevito Pull Request resolved: pytorch#10952 Differential Revision: D9544607 Pulled By: zou3519 fbshipit-source-id: 2b873f47e2377125e1f94eb1b310a95cda51476c
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
7 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.
Fixes #8560.
Unblocks #10715.
The assert (nDim <= uncompressedDims) was being triggered for a scalar
tensor because we compute nDim to be 1 for a scalar tensor but
uncompressedDim = 0.
This PR changes it so that we compute nDim to be 0 for a scalar tensor. This
works because indexing in a kernel depends on nDim. If nDim = 0, then
offset is always 0, which is what we want.
Some other (small) changes were necessary to make this work:
IndexType arr[0]so the codeguards against that
case when uncompressedDim == 0.
cc @apaszke @zdevito