[TensorExpr] Implement shape inference for TE. #41451
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
Since TE operates on a limited subset of ops with a well-defined semantics, we can easily infer shapes of intermediate and output tensors given shapes of the inputs. There is a couple of ops that are not yet supported in the shape inference, once we add them we could relax the shape info requirements in the TE fuser: currently it requires all values in the fusion group to have shapes known and we can change it to only inputs. [ghstack-poisoned]
ZolotukhinM
pushed a commit
that referenced
this pull request
Jul 15, 2020
Since TE operates on a limited subset of ops with a well-defined semantics, we can easily infer shapes of intermediate and output tensors given shapes of the inputs. There is a couple of ops that are not yet supported in the shape inference, once we add them we could relax the shape info requirements in the TE fuser: currently it requires all values in the fusion group to have shapes known and we can change it to only inputs. ghstack-source-id: 458c317 Pull Request resolved: #41451
facebook-github-bot
added
the
oncall: jit
💊 CI failures summary and remediationsAs of commit 4c3383f (more details on the Dr. CI page): 💚 💚 Looks good so far! There are no failures yet. 💚 💚 This comment was automatically generated by Dr. CI (expand for details).Follow this link to opt-out of these comments for your Pull Requests.Please report bugs/suggestions on the GitHub issue tracker or post in the (internal) Dr. CI Users group. This comment has been revised 28 times. |
Since TE operates on a limited subset of ops with a well-defined semantics, we can easily infer shapes of intermediate and output tensors given shapes of the inputs. There is a couple of ops that are not yet supported in the shape inference, once we add them we could relax the shape info requirements in the TE fuser: currently it requires all values in the fusion group to have shapes known and we can change it to only inputs. Differential Revision: [D22543470](https://our.internmc.facebook.com/intern/diff/D22543470) [ghstack-poisoned]
Since TE operates on a limited subset of ops with a well-defined semantics, we can easily infer shapes of intermediate and output tensors given shapes of the inputs. There is a couple of ops that are not yet supported in the shape inference, once we add them we could relax the shape info requirements in the TE fuser: currently it requires all values in the fusion group to have shapes known and we can change it to only inputs. Differential Revision: [D22543470](https://our.internmc.facebook.com/intern/diff/D22543470) [ghstack-poisoned]
This was referenced Jul 28, 2020
added 4 commits
July 28, 2020 17:34
Since TE operates on a limited subset of ops with a well-defined semantics, we can easily infer shapes of intermediate and output tensors given shapes of the inputs. There is a couple of ops that are not yet supported in the shape inference, once we add them we could relax the shape info requirements in the TE fuser: currently it requires all values in the fusion group to have shapes known and we can change it to only inputs. Differential Revision: [D22543470](https://our.internmc.facebook.com/intern/diff/D22543470) [ghstack-poisoned]
Since TE operates on a limited subset of ops with a well-defined semantics, we can easily infer shapes of intermediate and output tensors given shapes of the inputs. There is a couple of ops that are not yet supported in the shape inference, once we add them we could relax the shape info requirements in the TE fuser: currently it requires all values in the fusion group to have shapes known and we can change it to only inputs. Differential Revision: [D22543470](https://our.internmc.facebook.com/intern/diff/D22543470) [ghstack-poisoned]
Since TE operates on a limited subset of ops with a well-defined semantics, we can easily infer shapes of intermediate and output tensors given shapes of the inputs. There is a couple of ops that are not yet supported in the shape inference, once we add them we could relax the shape info requirements in the TE fuser: currently it requires all values in the fusion group to have shapes known and we can change it to only inputs. Differential Revision: [D22543470](https://our.internmc.facebook.com/intern/diff/D22543470) [ghstack-poisoned]
Since TE operates on a limited subset of ops with a well-defined semantics, we can easily infer shapes of intermediate and output tensors given shapes of the inputs. There is a couple of ops that are not yet supported in the shape inference, once we add them we could relax the shape info requirements in the TE fuser: currently it requires all values in the fusion group to have shapes known and we can change it to only inputs. Differential Revision: [D22543470](https://our.internmc.facebook.com/intern/diff/D22543470) [ghstack-poisoned]
eellison
approved these changes
Jul 31, 2020
There was a problem hiding this comment.
Accepting to unblock. We're going to need to think up a more comprehensive testing strategy before enabling TE Fuser on in master, which should hopefully flesh out any issues here. I also suspect a lot of this will need to be rethought and revisited if/when we enable TE with unknown dimensions.
@Krovatkin maybe you can take a look as well
torch/csrc/jit/tensorexpr/kernel.cpp
Outdated
| static std::pair<std::vector<ExprHandle>, bool> broadcastShapes( | ||
| std::vector<ExprHandle> TensorExprKernel::broadcastShapes( | ||
| std::vector<std::vector<ExprHandle>> shapes) { | ||
| bool seenBroadcast = false; |
added 2 commits
July 31, 2020 15:57
Since TE operates on a limited subset of ops with a well-defined semantics, we can easily infer shapes of intermediate and output tensors given shapes of the inputs. There is a couple of ops that are not yet supported in the shape inference, once we add them we could relax the shape info requirements in the TE fuser: currently it requires all values in the fusion group to have shapes known and we can change it to only inputs. Differential Revision: [D22543470](https://our.internmc.facebook.com/intern/diff/D22543470) [ghstack-poisoned]
Since TE operates on a limited subset of ops with a well-defined semantics, we can easily infer shapes of intermediate and output tensors given shapes of the inputs. There is a couple of ops that are not yet supported in the shape inference, once we add them we could relax the shape info requirements in the TE fuser: currently it requires all values in the fusion group to have shapes known and we can change it to only inputs. Differential Revision: [D22543470](https://our.internmc.facebook.com/intern/diff/D22543470) [ghstack-poisoned]
|
@ZolotukhinM merged this pull request in 2deccce. |
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.
Stack from ghstack:
Since TE operates on a limited subset of ops with a well-defined
semantics, we can easily infer shapes of intermediate and output tensors
given shapes of the inputs.
There is a couple of ops that are not yet supported in the shape
inference, once we add them we could relax the shape info requirements
in the TE fuser: currently it requires all values in the fusion group to
have shapes known and we can change it to only inputs.
Differential Revision: D22543470