fix op benchmark OOM issue #29794
Closed
fix op benchmark OOM issue #29794
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
|
This pull request was exported from Phabricator. Differential Revision: D18500103 |
Summary: Pull Request resolved: #29794 Before this diff, all tests of an operator are created at once before testing. Once an operator is benchmarked, the same process will move to the next operator and so on. The issue is that the number of tests of a single operator could be > 100 which can cause OOM issues. This diff avoids creating all the tests of an operator at once by using generators which creates/runs test one by one. Test Plan: ``` buck run //caffe2/benchmarks/operator_benchmark:benchmark_all_quantized_test -- --iterations 1 # ---------------------------------------- # PyTorch/Caffe2 Operator Micro-benchmarks # ---------------------------------------- # Tag : short # Benchmarking PyTorch: relu # Mode: Eager # Name: relu_dims(3,4,5)_contigFalse_inplaceFalse_dtypetorch.quint8 # Input: dims: (3, 4, 5), contig: False, inplace: False, dtype: torch.quint8 Forward Execution Time (us) : 52.493 # Benchmarking PyTorch: relu # Mode: Eager # Name: relu_dims(3,4,5)_contigFalse_inplaceFalse_dtypetorch.qint8 # Input: dims: (3, 4, 5), contig: False, inplace: False, dtype: torch.qint8 Forward Execution Time (us) : 44.945 ... Differential Revision: D18500103 fbshipit-source-id: 751168bab18de0939db1352534086fb7e299de00
|
This pull request was exported from Phabricator. Differential Revision: D18500103 |
|
This pull request has been merged in f3b1572. |
csarofeen
pushed a commit
to mruberry/pytorch
that referenced
this pull request
Nov 18, 2019
Summary: Pull Request resolved: pytorch#29794 Before this diff, all tests of an operator are created at once before testing. Once an operator is benchmarked, the same process will move to the next operator and so on. The issue is that the number of tests of a single operator could be > 100 which can cause OOM issues. This diff avoids creating all the tests of an operator at once by using generators which creates/runs test one by one. Test Plan: ``` buck run //caffe2/benchmarks/operator_benchmark:benchmark_all_quantized_test -- --iterations 1 # ---------------------------------------- # PyTorch/Caffe2 Operator Micro-benchmarks # ---------------------------------------- # Tag : short # Benchmarking PyTorch: relu # Mode: Eager # Name: relu_dims(3,4,5)_contigFalse_inplaceFalse_dtypetorch.quint8 # Input: dims: (3, 4, 5), contig: False, inplace: False, dtype: torch.quint8 Forward Execution Time (us) : 52.493 # Benchmarking PyTorch: relu # Mode: Eager # Name: relu_dims(3,4,5)_contigFalse_inplaceFalse_dtypetorch.qint8 # Input: dims: (3, 4, 5), contig: False, inplace: False, dtype: torch.qint8 Forward Execution Time (us) : 44.945 ... Reviewed By: hl475 Differential Revision: D18500103 fbshipit-source-id: 747c0ad0d302177da04da36e112c67f154115b6e
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 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.
Summary: Before this diff, all tests of an operator are created at once before testing. Once an operator is benchmark, the same process will move to the next operator and so on. The issue is that the number of tests of a single operator could be > 100 which can cause OOM issues. This diff avoids creating all the tests of an operator at once by using generators which creates/runs test one by one.
Test Plan: