Remove hacky_wrapper from VariableType and TraceType #44005
Closed
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
Previously, VariableType and TraceType kernels were still written in the legacy way, i.e. they took one TensorOptions argument instead of scattered dtype, layout, device, pin_memory, and they used hacky_wrapper to be callable. This caused a re-wrapping step. Calling into a Variable or Tracing kernel required taking the individual scattered arguments, packing them into a TensorOptions, and the kernel itself then likely re-dispatched, scattering those again into individual arguments. Now with this PR, variable and tracing kernels are written in the new way and no hacky_wrapper or rewrapping is needed for them. Differential Revision: [D23466042](https://our.internmc.facebook.com/intern/diff/D23466042/) [ghstack-poisoned]
smessmer
added a commit
that referenced
this pull request
Sep 2, 2020
Previously, VariableType and TraceType kernels were still written in the legacy way, i.e. they took one TensorOptions argument instead of scattered dtype, layout, device, pin_memory, and they used hacky_wrapper to be callable. This caused a re-wrapping step. Calling into a Variable or Tracing kernel required taking the individual scattered arguments, packing them into a TensorOptions, and the kernel itself then likely re-dispatched, scattering those again into individual arguments. Now with this PR, variable and tracing kernels are written in the new way and no hacky_wrapper or rewrapping is needed for them. Differential Revision: [D23466042](https://our.internmc.facebook.com/intern/diff/D23466042/) ghstack-source-id: 111210090 Pull Request resolved: #44005
Codecov Report
@@ Coverage Diff @@
## gh/smessmer/252/base #44005 +/- ##
=====================================================
Coverage 68.06% 68.06%
=====================================================
Files 393 393
Lines 50918 50918
=====================================================
+ Hits 34655 34657 +2
+ Misses 16263 16261 -2
Continue to review full report at Codecov.
|
Previously, VariableType and TraceType kernels were still written in the legacy way, i.e. they took one TensorOptions argument instead of scattered dtype, layout, device, pin_memory, and they used hacky_wrapper to be callable. Now with this PR, variable and tracing kernels are written in the new way and no hacky_wrapper is needed for them. Differential Revision: [D23466042](https://our.internmc.facebook.com/intern/diff/D23466042/) [ghstack-poisoned]
smessmer
added a commit
that referenced
this pull request
Sep 2, 2020
Pull Request resolved: #44005 Previously, VariableType and TraceType kernels were still written in the legacy way, i.e. they took one TensorOptions argument instead of scattered dtype, layout, device, pin_memory, and they used hacky_wrapper to be callable. Now with this PR, variable and tracing kernels are written in the new way and no hacky_wrapper is needed for them. ghstack-source-id: 111266814 Differential Revision: [D23466042](https://our.internmc.facebook.com/intern/diff/D23466042/)
💊 CI failures summary and remediationsAs of commit b68b323 (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 9 times. |
This was referenced Sep 2, 2020
Closed
ezyang
approved these changes
Sep 3, 2020
bhosmer
approved these changes
Sep 7, 2020
There was a problem hiding this comment.
A few minor things inline, plus one thing about tracing TensorOptions (I think this relates to a semantics issue you mentioned in chat, but I thought the answer was that we could scatter them into addInput()).
Also I'd add to the PR description that the modification is specifically for use_c10_dispatcher: full.
Note: I think both build failures come down to
Sep 02 19:19:41 -- Build files have been written to: /var/lib/jenkins/workspace/build_test_custom_build/predictor
Sep 02 19:19:41 + make
Sep 02 19:19:41 Scanning dependencies of target Predictor
Sep 02 19:19:41 [ 50%] Building CXX object CMakeFiles/Predictor.dir/predictor.cpp.o
Sep 02 19:19:41 [100%] Linking CXX executable Predictor
Sep 02 19:19:44 [100%] Built target Predictor
Sep 02 19:19:44 + run_predictor
Sep 02 19:19:44 + cd /var/lib/jenkins/workspace/build_test_custom_build/predictor
Sep 02 19:19:44 + ./Predictor /var/lib/jenkins/workspace/build_test_custom_build/MobileNetV2.pt
Sep 02 19:19:44 terminate called after throwing an instance of 'torch::jit::ErrorReport'
Sep 02 19:19:44 what():
Sep 02 19:19:44
Sep 02 19:19:44 aten::_convolution(Tensor input, Tensor weight, Tensor? bias, int[] stride, int[] padding, int[] dilation, bool transposed, int[] output_padding, int groups, bool benchmark, bool deterministic, bool cudnn_enabled) -> (Tensor):
Sep 02 19:19:44 Expected at most 12 arguments but found 13 positional arguments.
Previously, VariableType and TraceType kernels were still written in the legacy way, i.e. they took one TensorOptions argument instead of scattered dtype, layout, device, pin_memory, and they used hacky_wrapper to be callable. Now with this PR, variable and tracing kernels are written in the new way and no hacky_wrapper is needed for them. This only affects ops with `use_c10_dispatcher: full`. Differential Revision: [D23466042](https://our.internmc.facebook.com/intern/diff/D23466042/) [ghstack-poisoned]
Previously, VariableType and TraceType kernels were still written in the legacy way, i.e. they took one TensorOptions argument instead of scattered dtype, layout, device, pin_memory, and they used hacky_wrapper to be callable. Now with this PR, variable and tracing kernels are written in the new way and no hacky_wrapper is needed for them. This only affects ops with `use_c10_dispatcher: full`. Differential Revision: [D23466042](https://our.internmc.facebook.com/intern/diff/D23466042/) [ghstack-poisoned]
Previously, VariableType and TraceType kernels were still written in the legacy way, i.e. they took one TensorOptions argument instead of scattered dtype, layout, device, pin_memory, and they used hacky_wrapper to be callable. Now with this PR, variable and tracing kernels are written in the new way and no hacky_wrapper is needed for them. This only affects ops with `use_c10_dispatcher: full`. Differential Revision: [D23466042](https://our.internmc.facebook.com/intern/diff/D23466042/) [ghstack-poisoned]
|
This pull request has been merged in 043bd51. |
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:
Previously, VariableType and TraceType kernels were still written in the legacy way, i.e. they took one TensorOptions argument instead of scattered dtype, layout, device, pin_memory, and they used hacky_wrapper to be callable.
Now with this PR, variable and tracing kernels are written in the new way and no hacky_wrapper is needed for them.
This only affects ops with
use_c10_dispatcher: full.Differential Revision: D23466042