Allow caffe2-specific graph transformations for OperatorExportTypes.ONNX_ATEN_FALLBACK when BUILD_CAFFE2 is ON by thiagocrepaldi · Pull Request #67460 · pytorch/pytorch

thiagocrepaldi · 2021-10-28T19:03:40Z

The use of ATEN as a fallback operator during ONNX conversion is important for increasing operator coverage or even provide more efficient implementations over some ONNX ops.

Currently this feature is available through OperatorExportTypes.ONNX_ATEN_FALLBACK, but it also performs changes to
the graph that are runnable by Caffe2, only.

This PR introduces restricts caffe2-specific graph transformations for ONNX_ATEN_FALLBACK operator export type for when pytorch is built with caffe2 support (aka BUILD_CAFFE2=1 during build)

The first version of this PR introduced a new operator export type ONNX_ATEN__STRICT_FALLBACK, which essentially is the same as ONNX_ATEN_FALLBACK but without caffe2 transformations. It was preferred to not introduce a new operator export type, but to refine the existing aten fallback one

BC-breaking note

The global constant torch.onnx.PYTORCH_ONNX_CAFFE2_BUNDLE is removed in favor of a less visible torch.onnx._CAFFE2_ATEN_FALLBACK.

PYTORCH_ONNX_CAFFE2_BUNDLE is really a dead code flag always set to False. One alternative would be fixing it, but as #66658 disables Caffe2 build by default. Making a Caffe2 feature a private one seems to make more sense for future deprecation.

The method torch.onnx.export now defaults to `ONNX` when `operator_export_type` is not specified.

Previously torch.onnx.export's operator_export_type intended to default to ONNX_ATEN_FALLBACK when PYTORCH_ONNX_CAFFE2_BUNDLE was set, but it would never happend as PYTORCH_ONNX_CAFFE2_BUNDLE is always undefined

pytorch-probot · 2021-10-28T19:03:42Z

CI Flow Status

⚛️ CI Flow

Ruleset - Version: v1
Ruleset - File: https://github.com/thiagocrepaldi/pytorch/blob/6659836cf86b944a81d41242a0545a942cad25b8/.github/generated-ciflow-ruleset.json
PR ciflow labels: ciflow/default

Workflows	Labels (bold enabled)	Status
Triggered Workflows
linux-bionic-py3.6-clang9	`ciflow/all`, `ciflow/cpu`, `ciflow/default`, `ciflow/linux`, `ciflow/noarch`, `ciflow/xla`	✅ triggered
linux-vulkan-bionic-py3.6-clang9	`ciflow/all`, `ciflow/cpu`, `ciflow/default`, `ciflow/linux`, `ciflow/vulkan`	✅ triggered
linux-xenial-cuda11.3-py3.6-gcc7	`ciflow/all`, `ciflow/cuda`, `ciflow/default`, `ciflow/linux`	✅ triggered
linux-xenial-py3-clang5-mobile-build	`ciflow/all`, `ciflow/default`, `ciflow/linux`, `ciflow/mobile`	✅ triggered
linux-xenial-py3-clang5-mobile-custom-build-dynamic	`ciflow/all`, `ciflow/default`, `ciflow/linux`, `ciflow/mobile`	✅ triggered
linux-xenial-py3-clang5-mobile-custom-build-static	`ciflow/all`, `ciflow/default`, `ciflow/linux`, `ciflow/mobile`	✅ triggered
linux-xenial-py3.6-clang7-asan	`ciflow/all`, `ciflow/cpu`, `ciflow/default`, `ciflow/linux`, `ciflow/sanitizers`	✅ triggered
linux-xenial-py3.6-clang7-onnx	`ciflow/all`, `ciflow/cpu`, `ciflow/default`, `ciflow/linux`, `ciflow/onnx`	✅ triggered
linux-xenial-py3.6-gcc5.4	`ciflow/all`, `ciflow/cpu`, `ciflow/default`, `ciflow/linux`	✅ triggered
linux-xenial-py3.6-gcc7	`ciflow/all`, `ciflow/cpu`, `ciflow/default`, `ciflow/linux`	✅ triggered
linux-xenial-py3.6-gcc7-bazel-test	`ciflow/all`, `ciflow/bazel`, `ciflow/cpu`, `ciflow/default`, `ciflow/linux`	✅ triggered
pytorch-linux-xenial-py3-clang5-android-ndk-r19c-gradle-custom-build-single	`ciflow/all`, `ciflow/android`, `ciflow/cpu`, `ciflow/default`, `ciflow/linux`	✅ triggered
pytorch-linux-xenial-py3-clang5-android-ndk-r19c-gradle-custom-build-single-full-jit	`ciflow/all`, `ciflow/android`, `ciflow/cpu`, `ciflow/default`, `ciflow/linux`	✅ triggered
win-vs2019-cpu-py3	`ciflow/all`, `ciflow/cpu`, `ciflow/default`, `ciflow/win`	✅ triggered
win-vs2019-cuda11.3-py3	`ciflow/all`, `ciflow/cuda`, `ciflow/default`, `ciflow/win`	✅ triggered
Skipped Workflows
caffe2-linux-xenial-py3.6-gcc5.4	`ciflow/all`, `ciflow/cpu`, `ciflow/linux`	🚫 skipped
docker-builds	`ciflow/all`	🚫 skipped
ios-12-5-1-arm64	`ciflow/all`, `ciflow/ios`, `ciflow/macos`	🚫 skipped
ios-12-5-1-arm64-coreml	`ciflow/all`, `ciflow/ios`, `ciflow/macos`	🚫 skipped
ios-12-5-1-arm64-custom-ops	`ciflow/all`, `ciflow/ios`, `ciflow/macos`	🚫 skipped
ios-12-5-1-arm64-full-jit	`ciflow/all`, `ciflow/ios`, `ciflow/macos`	🚫 skipped
ios-12-5-1-arm64-metal	`ciflow/all`, `ciflow/ios`, `ciflow/macos`	🚫 skipped
ios-12-5-1-x86-64	`ciflow/all`, `ciflow/ios`, `ciflow/macos`	🚫 skipped
ios-12-5-1-x86-64-coreml	`ciflow/all`, `ciflow/ios`, `ciflow/macos`	🚫 skipped
ios-12-5-1-x86-64-full-jit	`ciflow/all`, `ciflow/ios`, `ciflow/macos`	🚫 skipped
libtorch-linux-xenial-cuda10.2-py3.6-gcc7	`ciflow/all`, `ciflow/cuda`, `ciflow/libtorch`, `ciflow/linux`	🚫 skipped
libtorch-linux-xenial-cuda11.3-py3.6-gcc7	`ciflow/all`, `ciflow/cuda`, `ciflow/libtorch`, `ciflow/linux`	🚫 skipped
linux-bionic-cuda10.2-py3.9-gcc7	`ciflow/all`, `ciflow/cuda`, `ciflow/linux`, `ciflow/slow`	🚫 skipped
linux-xenial-py3-clang5-mobile-code-analysis	`ciflow/all`, `ciflow/linux`, `ciflow/mobile`	🚫 skipped
parallelnative-linux-xenial-py3.6-gcc5.4	`ciflow/all`, `ciflow/cpu`, `ciflow/linux`	🚫 skipped
periodic-libtorch-linux-xenial-cuda11.1-py3.6-gcc7	`ciflow/all`, `ciflow/cuda`, `ciflow/libtorch`, `ciflow/linux`, `ciflow/scheduled`	🚫 skipped
periodic-linux-xenial-cuda10.2-py3-gcc7-slow-gradcheck	`ciflow/all`, `ciflow/cuda`, `ciflow/linux`, `ciflow/scheduled`, `ciflow/slow`, `ciflow/slow-gradcheck`	🚫 skipped
periodic-linux-xenial-cuda11.1-py3.6-gcc7	`ciflow/all`, `ciflow/cuda`, `ciflow/linux`, `ciflow/scheduled`	🚫 skipped
periodic-win-vs2019-cuda11.1-py3	`ciflow/all`, `ciflow/cuda`, `ciflow/scheduled`, `ciflow/win`	🚫 skipped

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/slow

For more information, please take a look at the CI Flow Wiki.

facebook-github-bot · 2021-10-28T19:03:45Z

🔗 Helpful links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/67460
📄 Preview docs built from this PR
📄 Preview C++ docs built from this PR
🔧 Opt-in to CIFlow to control what jobs run on your PRs

💊 CI failures summary and remediations

As of commit 6659836 (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).

Please report bugs/suggestions to the (internal) Dr. CI Users group.

Click here to manually regenerate this comment.

torch/onnx/symbolic_opset9.py

The use of ATEN as a fallback operator during ONNX conversion is important for increasing operator coverage or even provide more efficient implementations over some ONNX ops. Currently this feature is available through OperatorExportTypes.ONNX_ATEN_FALLBACK, but it also performs changes to the graph that are runnable by Caffe2, only. This PR introduces ONNX_ATEN_STRICT_FALLBACK as an alternative fallback for other backend implementations, such as ONNX Runtime that only support ONNX and ATEN operators. One alternative that it was considered was using `PYTORCH_ONNX_CAFFE2_BUNDLE` along with `ONNX_ATEN_FALLBACK` to introduce Caffe2-specific changes, but it seems Caffe2 is supported even when `PYTORCH_ONNX_CAFFE2_BUNDLE` is `False`.

…EN_FALLBACK

garymm

Thanks for this!

torch/onnx/symbolic_opset11.py

torch/onnx/utils.py

torch/csrc/jit/serialization/export.cpp

torch/csrc/onnx/init.cpp

torch/onnx/utils.py

garymm

Thanks!

torch/onnx/utils.py

…NNX_ATEN_FALLBACK when BUILD_CAFFE2 is ON (#67460) The use of ATEN as a fallback operator during ONNX conversion is important for increasing operator coverage or even provide more efficient implementations over some ONNX ops. Currently this feature is available through `OperatorExportTypes.ONNX_ATEN_FALLBACK`, but it also performs changes to the graph that are runnable by Caffe2, only. This PR introduces restricts caffe2-specific graph transformations for `ONNX_ATEN_FALLBACK` operator export type for when pytorch is built with caffe2 support (aka BUILD_CAFFE2=1 during build) The first version of this PR introduced a new operator export type `ONNX_ATEN__STRICT_FALLBACK`, which essentially is the same as `ONNX_ATEN_FALLBACK` but without caffe2 transformations. It was preferred to not introduce a new operator export type, but to refine the existing aten fallback one ## BC-breaking note ### The global constant `torch.onnx.PYTORCH_ONNX_CAFFE2_BUNDLE` is removed in favor of a less visible `torch.onnx._CAFFE2_ATEN_FALLBACK`. `PYTORCH_ONNX_CAFFE2_BUNDLE` is really a dead code flag always set to False. One alternative would be fixing it, but #66658 disables Caffe2 build by default. Making a Caffe2 feature a private one seems to make more sense for future deprecation. ### The method `torch.onnx.export` now defaults to ONNX when `operator_export_type` is not specified. Previously `torch.onnx.export's operator_export_type` intended to default to `ONNX_ATEN_FALLBACK` when `PYTORCH_ONNX_CAFFE2_BUNDLE` was set, but it would never happen as `PYTORCH_ONNX_CAFFE2_BUNDLE` is always undefined [ghstack-poisoned]

1. Make description of ONNX_ATEN_FALLBACK more accurate (after pytorch#67460). 2. Specify minimum and maximum values for opset_version. This is pretty important information and we should make users dig through source code to find it.

[ONNX] minor clarifications of docstrings 1. Make description of ONNX_ATEN_FALLBACK more accurate (after #67460). 2. Specify minimum and maximum values for opset_version. This is pretty important information and we should make users dig through source code to find it.

…s for OperatorExportTypes.ONNX_ATEN_FALLBACK when BUILD_CAFFE2 is ON (#67460)" The use of ATEN as a fallback operator during ONNX conversion is important for increasing operator coverage or even provide more efficient implementations over some ONNX ops. Currently this feature is available through `OperatorExportTypes.ONNX_ATEN_FALLBACK`, but it also performs changes to the graph that are runnable by Caffe2, only. This PR introduces restricts caffe2-specific graph transformations for `ONNX_ATEN_FALLBACK` operator export type for when pytorch is built with caffe2 support (aka BUILD_CAFFE2=1 during build) The first version of this PR introduced a new operator export type `ONNX_ATEN__STRICT_FALLBACK`, which essentially is the same as `ONNX_ATEN_FALLBACK` but without caffe2 transformations. It was preferred to not introduce a new operator export type, but to refine the existing aten fallback one ## BC-breaking note ### The global constant `torch.onnx.PYTORCH_ONNX_CAFFE2_BUNDLE` is removed in favor of a less visible `torch.onnx._CAFFE2_ATEN_FALLBACK`. `PYTORCH_ONNX_CAFFE2_BUNDLE` is really a dead code flag always set to False. One alternative would be fixing it, but #66658 disables Caffe2 build by default. Making a Caffe2 feature a private one seems to make more sense for future deprecation. ### The method `torch.onnx.export` now defaults to ONNX when `operator_export_type` is not specified. Previously `torch.onnx.export's operator_export_type` intended to default to `ONNX_ATEN_FALLBACK` when `PYTORCH_ONNX_CAFFE2_BUNDLE` was set, but it would never happen as `PYTORCH_ONNX_CAFFE2_BUNDLE` is always undefined Differential Revision: [D32483781](https://our.internmc.facebook.com/intern/diff/D32483781) [ghstack-poisoned]

…portTypes.ONNX_ATEN_FALLBACK when BUILD_CAFFE2 is ON (#67460)" The use of ATEN as a fallback operator during ONNX conversion is important for increasing operator coverage or even provide more efficient implementations over some ONNX ops. Currently this feature is available through `OperatorExportTypes.ONNX_ATEN_FALLBACK`, but it also performs changes to the graph that are runnable by Caffe2, only. This PR introduces restricts caffe2-specific graph transformations for `ONNX_ATEN_FALLBACK` operator export type for when pytorch is built with caffe2 support (aka BUILD_CAFFE2=1 during build) The first version of this PR introduced a new operator export type `ONNX_ATEN__STRICT_FALLBACK`, which essentially is the same as `ONNX_ATEN_FALLBACK` but without caffe2 transformations. It was preferred to not introduce a new operator export type, but to refine the existing aten fallback one ## BC-breaking note ### The global constant `torch.onnx.PYTORCH_ONNX_CAFFE2_BUNDLE` is removed in favor of a less visible `torch.onnx._CAFFE2_ATEN_FALLBACK`. `PYTORCH_ONNX_CAFFE2_BUNDLE` is really a dead code flag always set to False. One alternative would be fixing it, but #66658 disables Caffe2 build by default. Making a Caffe2 feature a private one seems to make more sense for future deprecation. ### The method `torch.onnx.export` now defaults to ONNX when `operator_export_type` is not specified. Previously `torch.onnx.export's operator_export_type` intended to default to `ONNX_ATEN_FALLBACK` when `PYTORCH_ONNX_CAFFE2_BUNDLE` was set, but it would never happen as `PYTORCH_ONNX_CAFFE2_BUNDLE` is always undefined Differential Revision: [D32483781](https://our.internmc.facebook.com/intern/diff/D32483781) [ghstack-poisoned]

[ONNX] minor clarifications of docstrings 1. Make description of ONNX_ATEN_FALLBACK more accurate (after #67460). 2. Specify minimum and maximum values for opset_version. This is pretty important information and we should make users dig through source code to find it. [ghstack-poisoned]

…NNX_ATEN_FALLBACK when BUILD_CAFFE2 is ON (pytorch#67460) The use of ATEN as a fallback operator during ONNX conversion is important for increasing operator coverage or even provide more efficient implementations over some ONNX ops. Currently this feature is available through `OperatorExportTypes.ONNX_ATEN_FALLBACK`, but it also performs changes to the graph that are runnable by Caffe2, only. This PR introduces restricts caffe2-specific graph transformations for `ONNX_ATEN_FALLBACK` operator export type for when pytorch is built with caffe2 support (aka BUILD_CAFFE2=1 during build) The first version of this PR introduced a new operator export type `ONNX_ATEN__STRICT_FALLBACK`, which essentially is the same as `ONNX_ATEN_FALLBACK` but without caffe2 transformations. It was preferred to not introduce a new operator export type, but to refine the existing aten fallback one ## BC-breaking note ### The global constant `torch.onnx.PYTORCH_ONNX_CAFFE2_BUNDLE` is removed in favor of a less visible `torch.onnx._CAFFE2_ATEN_FALLBACK`. `PYTORCH_ONNX_CAFFE2_BUNDLE` is really a dead code flag always set to False. One alternative would be fixing it, but pytorch#66658 disables Caffe2 build by default. Making a Caffe2 feature a private one seems to make more sense for future deprecation. ### The method `torch.onnx.export` now defaults to ONNX when `operator_export_type` is not specified. Previously `torch.onnx.export's operator_export_type` intended to default to `ONNX_ATEN_FALLBACK` when `PYTORCH_ONNX_CAFFE2_BUNDLE` was set, but it would never happen as `PYTORCH_ONNX_CAFFE2_BUNDLE` is always undefined ghstack-source-id: 8ae8ac0 Pull Request resolved: pytorch#68490

[ONNX] minor clarifications of docstrings 1. Make description of ONNX_ATEN_FALLBACK more accurate (after pytorch#67460). 2. Specify minimum and maximum values for opset_version. This is pretty important information and we should make users dig through source code to find it. ghstack-source-id: 6572ba2 Pull Request resolved: pytorch#69549

…s for OperatorExportTypes.ONNX_ATEN_FALLBACK when BUILD_CAFFE2 is ON (#67460)" The use of ATEN as a fallback operator during ONNX conversion is important for increasing operator coverage or even provide more efficient implementations over some ONNX ops. Currently this feature is available through `OperatorExportTypes.ONNX_ATEN_FALLBACK`, but it also performs changes to the graph that are runnable by Caffe2, only. This PR introduces restricts caffe2-specific graph transformations for `ONNX_ATEN_FALLBACK` operator export type for when pytorch is built with caffe2 support (aka BUILD_CAFFE2=1 during build) The first version of this PR introduced a new operator export type `ONNX_ATEN__STRICT_FALLBACK`, which essentially is the same as `ONNX_ATEN_FALLBACK` but without caffe2 transformations. It was preferred to not introduce a new operator export type, but to refine the existing aten fallback one ## BC-breaking note ### The global constant `torch.onnx.PYTORCH_ONNX_CAFFE2_BUNDLE` is removed in favor of a less visible `torch.onnx._CAFFE2_ATEN_FALLBACK`. `PYTORCH_ONNX_CAFFE2_BUNDLE` is really a dead code flag always set to False. One alternative would be fixing it, but #66658 disables Caffe2 build by default. Making a Caffe2 feature a private one seems to make more sense for future deprecation. ### The method `torch.onnx.export` now defaults to ONNX when `operator_export_type` is not specified. Previously `torch.onnx.export's operator_export_type` intended to default to `ONNX_ATEN_FALLBACK` when `PYTORCH_ONNX_CAFFE2_BUNDLE` was set, but it would never happen as `PYTORCH_ONNX_CAFFE2_BUNDLE` is always undefined Differential Revision: [D32483781](https://our.internmc.facebook.com/intern/diff/D32483781) [ghstack-poisoned]

…69260)" [ONNX] minor clarifications of docstrings 1. Make description of ONNX_ATEN_FALLBACK more accurate (after #67460). 2. Specify minimum and maximum values for opset_version. This is pretty important information and we should make users dig through source code to find it. Differential Revision: [D32994267](https://our.internmc.facebook.com/intern/diff/D32994267) [ghstack-poisoned]

[ONNX] minor clarifications of docstrings 1. Make description of ONNX_ATEN_FALLBACK more accurate (after #67460). 2. Specify minimum and maximum values for opset_version. This is pretty important information and we should make users dig through source code to find it. Differential Revision: [D32994267](https://our.internmc.facebook.com/intern/diff/D32994267) [ghstack-poisoned]

…portTypes.ONNX_ATEN_FALLBACK when BUILD_CAFFE2 is ON (#67460)" The use of ATEN as a fallback operator during ONNX conversion is important for increasing operator coverage or even provide more efficient implementations over some ONNX ops. Currently this feature is available through `OperatorExportTypes.ONNX_ATEN_FALLBACK`, but it also performs changes to the graph that are runnable by Caffe2, only. This PR introduces restricts caffe2-specific graph transformations for `ONNX_ATEN_FALLBACK` operator export type for when pytorch is built with caffe2 support (aka BUILD_CAFFE2=1 during build) The first version of this PR introduced a new operator export type `ONNX_ATEN__STRICT_FALLBACK`, which essentially is the same as `ONNX_ATEN_FALLBACK` but without caffe2 transformations. It was preferred to not introduce a new operator export type, but to refine the existing aten fallback one ## BC-breaking note ### The global constant `torch.onnx.PYTORCH_ONNX_CAFFE2_BUNDLE` is removed in favor of a less visible `torch.onnx._CAFFE2_ATEN_FALLBACK`. `PYTORCH_ONNX_CAFFE2_BUNDLE` is really a dead code flag always set to False. One alternative would be fixing it, but #66658 disables Caffe2 build by default. Making a Caffe2 feature a private one seems to make more sense for future deprecation. ### The method `torch.onnx.export` now defaults to ONNX when `operator_export_type` is not specified. Previously `torch.onnx.export's operator_export_type` intended to default to `ONNX_ATEN_FALLBACK` when `PYTORCH_ONNX_CAFFE2_BUNDLE` was set, but it would never happen as `PYTORCH_ONNX_CAFFE2_BUNDLE` is always undefined Differential Revision: [D32483781](https://our.internmc.facebook.com/intern/diff/D32483781) [ghstack-poisoned]

…s for OperatorExportTypes.ONNX_ATEN_FALLBACK when BUILD_CAFFE2 is ON (#67460)" The use of ATEN as a fallback operator during ONNX conversion is important for increasing operator coverage or even provide more efficient implementations over some ONNX ops. Currently this feature is available through `OperatorExportTypes.ONNX_ATEN_FALLBACK`, but it also performs changes to the graph that are runnable by Caffe2, only. This PR introduces restricts caffe2-specific graph transformations for `ONNX_ATEN_FALLBACK` operator export type for when pytorch is built with caffe2 support (aka BUILD_CAFFE2=1 during build) The first version of this PR introduced a new operator export type `ONNX_ATEN__STRICT_FALLBACK`, which essentially is the same as `ONNX_ATEN_FALLBACK` but without caffe2 transformations. It was preferred to not introduce a new operator export type, but to refine the existing aten fallback one ## BC-breaking note ### The global constant `torch.onnx.PYTORCH_ONNX_CAFFE2_BUNDLE` is removed in favor of a less visible `torch.onnx._CAFFE2_ATEN_FALLBACK`. `PYTORCH_ONNX_CAFFE2_BUNDLE` is really a dead code flag always set to False. One alternative would be fixing it, but #66658 disables Caffe2 build by default. Making a Caffe2 feature a private one seems to make more sense for future deprecation. ### The method `torch.onnx.export` now defaults to ONNX when `operator_export_type` is not specified. Previously `torch.onnx.export's operator_export_type` intended to default to `ONNX_ATEN_FALLBACK` when `PYTORCH_ONNX_CAFFE2_BUNDLE` was set, but it would never happen as `PYTORCH_ONNX_CAFFE2_BUNDLE` is always undefined Differential Revision: [D32483781](https://our.internmc.facebook.com/intern/diff/D32483781) [ghstack-poisoned]

…portTypes.ONNX_ATEN_FALLBACK when BUILD_CAFFE2 is ON (#67460)" The use of ATEN as a fallback operator during ONNX conversion is important for increasing operator coverage or even provide more efficient implementations over some ONNX ops. Currently this feature is available through `OperatorExportTypes.ONNX_ATEN_FALLBACK`, but it also performs changes to the graph that are runnable by Caffe2, only. This PR introduces restricts caffe2-specific graph transformations for `ONNX_ATEN_FALLBACK` operator export type for when pytorch is built with caffe2 support (aka BUILD_CAFFE2=1 during build) The first version of this PR introduced a new operator export type `ONNX_ATEN__STRICT_FALLBACK`, which essentially is the same as `ONNX_ATEN_FALLBACK` but without caffe2 transformations. It was preferred to not introduce a new operator export type, but to refine the existing aten fallback one ## BC-breaking note ### The global constant `torch.onnx.PYTORCH_ONNX_CAFFE2_BUNDLE` is removed in favor of a less visible `torch.onnx._CAFFE2_ATEN_FALLBACK`. `PYTORCH_ONNX_CAFFE2_BUNDLE` is really a dead code flag always set to False. One alternative would be fixing it, but #66658 disables Caffe2 build by default. Making a Caffe2 feature a private one seems to make more sense for future deprecation. ### The method `torch.onnx.export` now defaults to ONNX when `operator_export_type` is not specified. Previously `torch.onnx.export's operator_export_type` intended to default to `ONNX_ATEN_FALLBACK` when `PYTORCH_ONNX_CAFFE2_BUNDLE` was set, but it would never happen as `PYTORCH_ONNX_CAFFE2_BUNDLE` is always undefined Differential Revision: [D32483781](https://our.internmc.facebook.com/intern/diff/D32483781) [ghstack-poisoned]

…69260)" [ONNX] minor clarifications of docstrings 1. Make description of ONNX_ATEN_FALLBACK more accurate (after #67460). 2. Specify minimum and maximum values for opset_version. This is pretty important information and we should make users dig through source code to find it. Differential Revision: [D32994267](https://our.internmc.facebook.com/intern/diff/D32994267) [ghstack-poisoned]

[ONNX] minor clarifications of docstrings 1. Make description of ONNX_ATEN_FALLBACK more accurate (after #67460). 2. Specify minimum and maximum values for opset_version. This is pretty important information and we should make users dig through source code to find it. Differential Revision: [D32994267](https://our.internmc.facebook.com/intern/diff/D32994267) [ghstack-poisoned]

Summary: Pull Request resolved: #69549 [ONNX] minor clarifications of docstrings 1. Make description of ONNX_ATEN_FALLBACK more accurate (after #67460). 2. Specify minimum and maximum values for opset_version. This is pretty important information and we should make users dig through source code to find it. Test Plan: Imported from OSS Reviewed By: malfet Differential Revision: D32994267 Pulled By: msaroufim fbshipit-source-id: ba641404107baa23506d337eca742fc1fe9f0772

…NNX_ATEN_FALLBACK when BUILD_CAFFE2 is ON (pytorch#67460) The use of ATEN as a fallback operator during ONNX conversion is important for increasing operator coverage or even provide more efficient implementations over some ONNX ops. Currently this feature is available through `OperatorExportTypes.ONNX_ATEN_FALLBACK`, but it also performs changes to the graph that are runnable by Caffe2, only. This PR introduces restricts caffe2-specific graph transformations for `ONNX_ATEN_FALLBACK` operator export type for when pytorch is built with caffe2 support (aka BUILD_CAFFE2=1 during build) The first version of this PR introduced a new operator export type `ONNX_ATEN__STRICT_FALLBACK`, which essentially is the same as `ONNX_ATEN_FALLBACK` but without caffe2 transformations. It was preferred to not introduce a new operator export type, but to refine the existing aten fallback one ## BC-breaking note ### The global constant `torch.onnx.PYTORCH_ONNX_CAFFE2_BUNDLE` is removed in favor of a less visible `torch.onnx._CAFFE2_ATEN_FALLBACK`. `PYTORCH_ONNX_CAFFE2_BUNDLE` is really a dead code flag always set to False. One alternative would be fixing it, but pytorch#66658 disables Caffe2 build by default. Making a Caffe2 feature a private one seems to make more sense for future deprecation. ### The method `torch.onnx.export` now defaults to ONNX when `operator_export_type` is not specified. Previously `torch.onnx.export's operator_export_type` intended to default to `ONNX_ATEN_FALLBACK` when `PYTORCH_ONNX_CAFFE2_BUNDLE` was set, but it would never happen as `PYTORCH_ONNX_CAFFE2_BUNDLE` is always undefined ghstack-source-id: 8ae8ac0 Pull Request resolved: pytorch#68490

pytorch-probot bot added the ciflow/default label Oct 28, 2021

facebook-github-bot added the cla signed label Oct 28, 2021

facebook-github-bot added the oncall: jit Add this issue/PR to JIT oncall triage queue label Oct 28, 2021

pytorchbot added the open source label Oct 28, 2021

thiagocrepaldi force-pushed the thiagofc/add-onnx-aten-strict-fallback branch from 62624d3 to aecd2e6 Compare October 28, 2021 19:36

thiagocrepaldi changed the title ~~[WIP] Add ONNX_ATEN_FALLBACK as new torch::onnx::OperatorExportTypes~~ Add ONNX_ATEN_FALLBACK as new torch::onnx::OperatorExportTypes Oct 29, 2021

thiagocrepaldi marked this pull request as ready for review October 29, 2021 19:26

thiagocrepaldi requested review from BowenBao, neginraoof and shubhambhokare1 as code owners October 29, 2021 19:26

BowenBao reviewed Oct 29, 2021

View reviewed changes

torch/onnx/symbolic_opset9.py Outdated Show resolved Hide resolved

thiagocrepaldi requested a review from BowenBao November 1, 2021 20:27

thiagocrepaldi force-pushed the thiagofc/add-onnx-aten-strict-fallback branch from 7af56fc to 8acc3d3 Compare November 2, 2021 20:41

thiagocrepaldi requested a review from garymm November 2, 2021 20:41

thiagocrepaldi changed the title ~~Add ONNX_ATEN_FALLBACK as new torch::onnx::OperatorExportTypes~~ Restrict caffe2-specific changes for torch::onnx::OperatorExportTypes.ONNX_ATEN_FALLBACK when BUILD_CAFFE2 is ON Nov 2, 2021

thiagocrepaldi changed the title ~~Restrict caffe2-specific changes for torch::onnx::OperatorExportTypes.ONNX_ATEN_FALLBACK when BUILD_CAFFE2 is ON~~ Allow caffe2-specific graph transformations for OperatorExportTypes.ONNX_ATEN_FALLBACK when BUILD_CAFFE2 is ON Nov 2, 2021

thiagocrepaldi force-pushed the thiagofc/add-onnx-aten-strict-fallback branch 2 times, most recently from a6aa2a5 to 0440fb9 Compare November 3, 2021 15:42

BowenBao force-pushed the onnx_ms_1 branch from 7c78a1f to 1a91e0b Compare November 4, 2021 18:41

Thiago Crepaldi added 2 commits November 4, 2021 12:15

Combine ONNX_ATEN_FALLBACK and ONNX_ATEN_STRICT_FALLBACK into ONNX_AT…

3f86347

…EN_FALLBACK

thiagocrepaldi force-pushed the thiagofc/add-onnx-aten-strict-fallback branch from 0440fb9 to 3f86347 Compare November 4, 2021 19:16

garymm requested changes Nov 4, 2021

View reviewed changes

thiagocrepaldi requested a review from garymm November 5, 2021 16:44

thiagocrepaldi force-pushed the thiagofc/add-onnx-aten-strict-fallback branch from a34e12f to b99931a Compare November 5, 2021 16:53

garymm requested changes Nov 5, 2021

View reviewed changes

torch/csrc/onnx/init.cpp Show resolved Hide resolved

torch/csrc/onnx/init.cpp Outdated Show resolved Hide resolved

torch/onnx/utils.py Outdated Show resolved Hide resolved

thiagocrepaldi force-pushed the thiagofc/add-onnx-aten-strict-fallback branch from b99931a to 38ffc5c Compare November 5, 2021 17:29

thiagocrepaldi requested a review from garymm November 5, 2021 17:30

thiagocrepaldi force-pushed the thiagofc/add-onnx-aten-strict-fallback branch from 5ac32b3 to c4d57ca Compare November 5, 2021 18:06

thiagocrepaldi requested a review from garymm November 5, 2021 18:08

garymm approved these changes Nov 5, 2021

View reviewed changes

torch/onnx/utils.py Outdated Show resolved Hide resolved

thiagocrepaldi force-pushed the thiagofc/add-onnx-aten-strict-fallback branch from c4d57ca to fef50d3 Compare November 5, 2021 18:24

Address comments

6659836

thiagocrepaldi force-pushed the thiagofc/add-onnx-aten-strict-fallback branch from fef50d3 to 6659836 Compare November 5, 2021 18:56

thiagocrepaldi merged commit a14c317 into pytorch:onnx_ms_1 Nov 8, 2021

garymm mentioned this pull request Dec 1, 2021

[ONNX] minor clarifications of docstrings #69260

Merged

BowenBao mentioned this pull request Dec 7, 2021

[ONNX] minor clarifications of docstrings (#69260) #69549

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Allow caffe2-specific graph transformations for OperatorExportTypes.ONNX_ATEN_FALLBACK when BUILD_CAFFE2 is ON#67460

Allow caffe2-specific graph transformations for OperatorExportTypes.ONNX_ATEN_FALLBACK when BUILD_CAFFE2 is ON#67460
thiagocrepaldi merged 3 commits intopytorch:onnx_ms_1from
thiagocrepaldi:thiagofc/add-onnx-aten-strict-fallback

thiagocrepaldi commented Oct 28, 2021 •

edited

Loading

Uh oh!

pytorch-probot bot commented Oct 28, 2021 •

edited

Loading

⚛️ CI Flow

Uh oh!

facebook-github-bot commented Oct 28, 2021 •

edited

Loading

Uh oh!

Uh oh!

garymm left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

garymm left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Conversation

thiagocrepaldi commented Oct 28, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

BC-breaking note

The global constant torch.onnx.PYTORCH_ONNX_CAFFE2_BUNDLE is removed in favor of a less visible torch.onnx._CAFFE2_ATEN_FALLBACK.

The method torch.onnx.export now defaults to ONNX when operator_export_type is not specified.

Uh oh!

pytorch-probot bot commented Oct 28, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

⚛️ CI Flow

Uh oh!

facebook-github-bot commented Oct 28, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🔗 Helpful links

💊 CI failures summary and remediations

Uh oh!

Uh oh!

garymm left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

garymm left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

thiagocrepaldi commented Oct 28, 2021 •

edited

Loading

The method torch.onnx.export now defaults to `ONNX` when `operator_export_type` is not specified.

pytorch-probot bot commented Oct 28, 2021 •

edited

Loading

facebook-github-bot commented Oct 28, 2021 •

edited

Loading