[FX] fuse permute021 linear pass for trt lowering #66362
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
CI Flow Status⚛️ CI FlowRuleset - Version:
You can add a comment to the PR and tag @pytorchbot with the following commands: # ciflow rerun, "ciflow/default" will always be added automatically
@pytorchbot ciflow rerun
# ciflow rerun with additional labels "-l <ciflow/label_name>", which is equivalent to adding these labels manually and trigger the rerun
@pytorchbot ciflow rerun -l ciflow/scheduled -l ciflow/slowFor more information, please take a look at the CI Flow Wiki. |
🔗 Helpful links
💊 CI failures summary and remediationsAs of commit 5c5fee8 (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 to the (internal) Dr. CI Users group. |
|
This pull request was exported from Phabricator. Differential Revision: D31525307 |
1 similar comment
|
This pull request was exported from Phabricator. Differential Revision: D31525307 |
yinghai
force-pushed
the
export-D31525307
branch
from
a77562d to
601e659
Compare
yinghai
force-pushed
the
export-D31525307
branch
from
601e659 to
274b612
Compare
|
This pull request was exported from Phabricator. Differential Revision: D31525307 |
1 similar comment
|
This pull request was exported from Phabricator. Differential Revision: D31525307 |
yinghai
force-pushed
the
export-D31525307
branch
from
274b612 to
e2ddc22
Compare
Summary: Pull Request resolved: pytorch#66362 In general we cannot rely on Permute021Linear being kept as is before lowering phase before our transformation could have traced through this module. A acc based fx pass is more reliable to recover the perf. Test Plan: ``` buck run mode/opt -c python.package_style=inplace -c fbcode.nvcc_arch=a100 //hpc/new/models/ads/benchmarks:ads_dense_benchmark -- over-arch --model-version=23x_3tb --batch-size=2048 OverArch, PyTorch, FP16, BS: 2048, TFLOP/s: 53.22, Time per iter: 14.46ms, QPS: 141629.45 OverArch, TensorRT, FP16, BS: 2048, TFLOP/s: 92.20, Time per iter: 8.35ms, QPS: 245354.15 ``` Unittest: ``` buck test mode/dev-nosan caffe2/torch/fb/fx2trt:test_fuse_permute_linear_trt ``` Reviewed By: jianyuh, wushirong, 842974287 Differential Revision: D31525307 fbshipit-source-id: 6a67991125792110c1aefd054bff1658a78016a0
yinghai
force-pushed
the
export-D31525307
branch
from
e2ddc22 to
5c5fee8
Compare
|
This pull request was exported from Phabricator. Differential Revision: D31525307 |
yinghai
pushed a commit
to yinghai/pytorch
that referenced
this pull request
Oct 13, 2021
…intermediate node (pytorch#66472) Summary: Pull Request resolved: pytorch#66472 A follow up of pytorch#66362. Same fix. Test Plan: ``` buck test mode/dev-nosan caffe2/torch/fb/fx2trt:test_fuse_permute_matmul_trt buck test mode/dev-nosan caffe2/torch/fb/fx2trt:test_fuse_permute_linear_trt ``` Reviewed By: wushirong, 842974287 Differential Revision: D31567662 fbshipit-source-id: 43c366dea412346b4cecc7b17ba0f85ca98ac34a
facebook-github-bot
pushed a commit
that referenced
this pull request
Oct 13, 2021
…intermediate node (#66472) Summary: Pull Request resolved: #66472 A follow up of #66362. Same fix. Test Plan: ``` buck test mode/dev-nosan caffe2/torch/fb/fx2trt:test_fuse_permute_matmul_trt buck test mode/dev-nosan caffe2/torch/fb/fx2trt:test_fuse_permute_linear_trt ``` Reviewed By: wushirong, 842974287 Differential Revision: D31567662 fbshipit-source-id: 2c9e6a138fc31996d790fd4d79e0bf931507fc99
wconstab
pushed a commit
that referenced
this pull request
Oct 20, 2021
Summary: Pull Request resolved: #66362 In general we cannot rely on Permute021Linear being kept as is before lowering phase before our transformation could have traced through this module. A acc based fx pass is more reliable to recover the perf. Test Plan: ``` buck run mode/opt -c python.package_style=inplace -c fbcode.nvcc_arch=a100 //hpc/new/models/ads/benchmarks:ads_dense_benchmark -- over-arch --model-version=23x_3tb --batch-size=2048 OverArch, PyTorch, FP16, BS: 2048, TFLOP/s: 53.22, Time per iter: 14.46ms, QPS: 141629.45 OverArch, TensorRT, FP16, BS: 2048, TFLOP/s: 92.20, Time per iter: 8.35ms, QPS: 245354.15 ``` Unittest: ``` buck test mode/dev-nosan caffe2/torch/fb/fx2trt:test_fuse_permute_linear_trt ``` Reviewed By: jianyuh, wushirong, 842974287 Differential Revision: D31525307 fbshipit-source-id: b472a8c277aa4d156d933d6a5abec091133f22c5
wconstab
pushed a commit
that referenced
this pull request
Oct 20, 2021
…intermediate node (#66472) Summary: Pull Request resolved: #66472 A follow up of #66362. Same fix. Test Plan: ``` buck test mode/dev-nosan caffe2/torch/fb/fx2trt:test_fuse_permute_matmul_trt buck test mode/dev-nosan caffe2/torch/fb/fx2trt:test_fuse_permute_linear_trt ``` Reviewed By: wushirong, 842974287 Differential Revision: D31567662 fbshipit-source-id: 2c9e6a138fc31996d790fd4d79e0bf931507fc99
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 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: In general we cannot rely on Permute021Linear being kept as is before lowering phase before our transformation could have traced through this module. A acc based fx pass is more reliable to recover the perf.
Test Plan:
Unittest:
Differential Revision: D31525307