Skip to content

Add new ways of matching nodes with schemas in the JIT#9567

Closed
apaszke wants to merge 2 commits intopytorch:masterfrom
apaszke:jit_pattern_match
Closed

Add new ways of matching nodes with schemas in the JIT#9567
apaszke wants to merge 2 commits intopytorch:masterfrom
apaszke:jit_pattern_match

Conversation

@apaszke
Copy link
Contributor

@apaszke apaszke commented Jul 18, 2018

REVIEW LAST COMMIT ONLY

As discussed in our yesterday's meeting. Nodes can be now matched to particular overloads using the matches(...) function:

n->matches("aten::type_as(Tensor self, Tensor other) -> Tensor")

This also changes the shape prop and peephole passes to use those functions for matching. This fixes a few bugs, makes them much more robust, and prepares us for removal of attributes.

@zdevito

zdevito
zdevito approved these changes Jul 19, 2018
View reviewed changes
Copy link
Contributor

@zdevito zdevito left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I have some suggestions but this generally looks good. We should try to find ways to make obtaining constant values more concise because the n->get<type>(attr).value() pattern is pretty verbose.

torch/csrc/jit/function_schema.h Outdated

This comment was marked as off-topic.

Sign in to view

This comment was marked as off-topic.

Sign in to view

This comment was marked as off-topic.

Sign in to view

torch/csrc/jit/ir.h Outdated

This comment was marked as off-topic.

Sign in to view

torch/csrc/jit/ir.h Outdated

This comment was marked as off-topic.

Sign in to view

This comment was marked as off-topic.

Sign in to view

torch/csrc/jit/ir.cpp Outdated

This comment was marked as off-topic.

Sign in to view

This comment was marked as off-topic.

Sign in to view

torch/csrc/jit/operator.cpp Outdated

This comment was marked as off-topic.

Sign in to view

torch/csrc/jit/operator.cpp Outdated

This comment was marked as off-topic.

Sign in to view

torch/csrc/jit/operator.cpp Outdated

This comment was marked as off-topic.

Sign in to view

This comment was marked as off-topic.

Sign in to view

torch/csrc/jit/passes/peephole.cpp Outdated

This comment was marked as off-topic.

Sign in to view

This comment was marked as off-topic.

Sign in to view

@apaszke apaszke force-pushed the jit_pattern_match branch from ee9c001 to 4bb4044 Compare July 20, 2018 18:11
facebook-github-bot
facebook-github-bot reviewed Jul 20, 2018
View reviewed changes
Copy link
Contributor

@facebook-github-bot facebook-github-bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@apaszke has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

zdevito pushed a commit to zdevito/ATen that referenced this pull request Jul 22, 2018
@apaszke @facebook-github-bot
Add new ways of matching nodes with schemas in the JIT (#9567)
325c013 
Summary:
**REVIEW LAST COMMIT ONLY**

As discussed in our yesterday's meeting. Nodes can be now matched to particular overloads using the `matches(...)` function:
```cpp
n->matches("aten::type_as(Tensor self, Tensor other) -> Tensor")
```

This also changes the shape prop and peephole passes to use those functions for matching. This fixes a few bugs, makes them much more robust, and prepares us for removal of attributes.

zdevito
Pull Request resolved: pytorch/pytorch#9567

Reviewed By: zdevito

Differential Revision: D8938482

Pulled By: apaszke

fbshipit-source-id: eb2382eeeae99692aada2d78d5d0c87c8ef1545e
@ezyang
Copy link
Contributor

ezyang commented Jul 22, 2018

Hey @apaszke, does this error message ring a bell for you?

RuntimeError: 
Schema not found for node. File a bug report.
Node: %277 : Float(4!, 7128) = aten::expand[size=[4, 7128], implicit=1](%52), scope: MultiTaskModel/Sequential/Linear[1]
Input types:Float(7128)
candidates were:
:
torch/nn/functional.py(1024): linear
torch/nn/modules/linear.py(55): forward
torch/nn/modules/module.py(465): _slow_forward
torch/nn/modules/module.py(475): __call__
torch/nn/modules/container.py(91): forward
torch/nn/modules/module.py(465): _slow_forward
torch/nn/modules/mo

jramseyer pushed a commit to jramseyer/pytorch that referenced this pull request Jul 30, 2018
@apaszke @jramseyer
Add new ways of matching nodes with schemas in the JIT (pytorch#9567)
e1ac209 
Summary:
**REVIEW LAST COMMIT ONLY**

As discussed in our yesterday's meeting. Nodes can be now matched to particular overloads using the `matches(...)` function:
```cpp
n->matches("aten::type_as(Tensor self, Tensor other) -> Tensor")
```

This also changes the shape prop and peephole passes to use those functions for matching. This fixes a few bugs, makes them much more robust, and prepares us for removal of attributes.

zdevito
Pull Request resolved: pytorch#9567

Reviewed By: zdevito

Differential Revision: D8938482

Pulled By: apaszke

fbshipit-source-id: eb2382eeeae99692aada2d78d5d0c87c8ef1545e
goodlux pushed a commit to goodlux/pytorch that referenced this pull request Aug 15, 2018
@apaszke
Add new ways of matching nodes with schemas in the JIT (pytorch#9567)
1d0a57b 
Summary:
**REVIEW LAST COMMIT ONLY**

As discussed in our yesterday's meeting. Nodes can be now matched to particular overloads using the `matches(...)` function:
```cpp
n->matches("aten::type_as(Tensor self, Tensor other) -> Tensor")
```

This also changes the shape prop and peephole passes to use those functions for matching. This fixes a few bugs, makes them much more robust, and prepares us for removal of attributes.

zdevito
Pull Request resolved: pytorch#9567

Reviewed By: zdevito

Differential Revision: D8938482

Pulled By: apaszke

fbshipit-source-id: eb2382eeeae99692aada2d78d5d0c87c8ef1545e
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@facebook-github-bot facebook-github-bot facebook-github-bot left review comments

@zdevito zdevito zdevito approved these changes

@colesbury colesbury Awaiting requested review from colesbury

@ezyang ezyang Awaiting requested review from ezyang

@gchanan gchanan Awaiting requested review from gchanan

@soumith soumith Awaiting requested review from soumith

Assignees

No one assigned

Labels

open source

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

4 participants

@apaszke @ezyang @zdevito @facebook-github-bot