Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[TensorRT EP] Add new provider option to exclude nodes from running on TRT #22681

Merged
merged 20 commits into from
Nov 13, 2024

Conversation

chilo-ms
Copy link
Contributor

@chilo-ms chilo-ms commented Nov 1, 2024

Add new provider option trt_op_types_to_exclude:

  • User can provide op type list to be excluded from running on TRT
  • e.g. trt_op_types_to_exclude="MaxPool"

There is a known performance issue with the DDS ops (NonMaxSuppression, NonZero and RoiAlign) from TRT versions 10.0 to 10.7. TRT EP excludes DDS ops from running on TRT by default, user can override default value with empty string to include all ops.

@chilo-ms chilo-ms requested a review from jywu-msft November 1, 2024 01:11
Copy link
Contributor

@github-actions github-actions bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You can commit the suggested changes from lintrunner.

Copy link
Contributor

@github-actions github-actions bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You can commit the suggested changes from lintrunner.

Copy link
Contributor

@github-actions github-actions bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You can commit the suggested changes from lintrunner.

onnxruntime/python/onnxruntime_pybind_state.cc Outdated Show resolved Hide resolved
Copy link
Contributor

@github-actions github-actions bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You can commit the suggested changes from lintrunner.

Copy link
Contributor

@github-actions github-actions bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You can commit the suggested changes from lintrunner.

onnxruntime/test/providers/tensorrt/tensorrt_basic_test.cc Outdated Show resolved Hide resolved
Copy link
Contributor

@github-actions github-actions bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You can commit the suggested changes from lintrunner.

onnxruntime/test/providers/tensorrt/tensorrt_basic_test.cc Outdated Show resolved Hide resolved
Copy link
Contributor

@github-actions github-actions bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You can commit the suggested changes from lintrunner.

Copy link
Contributor

@github-actions github-actions bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You can commit the suggested changes from lintrunner.

@sophies927 sophies927 added release:1.20.1 triage:approved Approved for cherrypicks for release labels Nov 13, 2024
@chilo-ms chilo-ms merged commit fa4cbcd into main Nov 13, 2024
93 checks passed
@chilo-ms chilo-ms deleted the chi/exclude_nodes_from_trt_ep branch November 13, 2024 19:34
yf711 pushed a commit that referenced this pull request Nov 14, 2024
…n TRT (#22681)

Add new provider option `trt_op_types_to_exclude`:
- User can provide op type list to be excluded from running on TRT
- e.g. `trt_op_types_to_exclude="MaxPool"`

There is a known performance issue with the DDS ops (NonMaxSuppression,
NonZero and RoiAlign) from TRT versions 10.0 to 10.7. TRT EP excludes
DDS ops from running on TRT by default, user can override default value
with empty string to include all ops.
yf711 added a commit that referenced this pull request Nov 18, 2024
chilo-ms added a commit that referenced this pull request Nov 18, 2024
yf711 pushed a commit that referenced this pull request Nov 18, 2024
TRT EP excludes DDS ops from running on TRT and doesn't allow users to
change.
This PR is for ORT 1.20.1 patch release.

We will have better solution to add a new provider option to exclude
specific ops, similar to following:
#22863
#22681
yf711 pushed a commit that referenced this pull request Nov 18, 2024
TRT EP excludes DDS ops from running on TRT and doesn't allow users to
change.
This PR is for ORT 1.20.1 patch release.

We will have better solution to add a new provider option to exclude
specific ops, similar to following:
#22863
#22681
@yf711 yf711 removed triage:approved Approved for cherrypicks for release release:1.20.1 labels Nov 18, 2024
ishwar-raut1 pushed a commit to ishwar-raut1/onnxruntime that referenced this pull request Nov 19, 2024
…n TRT (microsoft#22681)

Add new provider option `trt_op_types_to_exclude`:
- User can provide op type list to be excluded from running on TRT
- e.g. `trt_op_types_to_exclude="MaxPool"`

There is a known performance issue with the DDS ops (NonMaxSuppression,
NonZero and RoiAlign) from TRT versions 10.0 to 10.7. TRT EP excludes
DDS ops from running on TRT by default, user can override default value
with empty string to include all ops.
chilo-ms added a commit that referenced this pull request Nov 19, 2024
…unning on TRT" (#22878)

- Revert #22681
- But still implicitly exclude DDS ops for TRT 10. Will later provide
better PR to add trt_op_types_to_exclude provider option.
mszhanyi pushed a commit that referenced this pull request Nov 22, 2024
…unning on TRT" (#22878)

- Revert #22681
- But still implicitly exclude DDS ops for TRT 10. Will later provide
better PR to add trt_op_types_to_exclude provider option.
guschmue pushed a commit that referenced this pull request Dec 2, 2024
…n TRT (#22681)

Add new provider option `trt_op_types_to_exclude`:
- User can provide op type list to be excluded from running on TRT
- e.g. `trt_op_types_to_exclude="MaxPool"`

There is a known performance issue with the DDS ops (NonMaxSuppression,
NonZero and RoiAlign) from TRT versions 10.0 to 10.7. TRT EP excludes
DDS ops from running on TRT by default, user can override default value
with empty string to include all ops.
guschmue pushed a commit that referenced this pull request Dec 2, 2024
…unning on TRT" (#22878)

- Revert #22681
- But still implicitly exclude DDS ops for TRT 10. Will later provide
better PR to add trt_op_types_to_exclude provider option.
ankitm3k pushed a commit to intel/onnxruntime that referenced this pull request Dec 11, 2024
…n TRT (microsoft#22681)

Add new provider option `trt_op_types_to_exclude`:
- User can provide op type list to be excluded from running on TRT
- e.g. `trt_op_types_to_exclude="MaxPool"`

There is a known performance issue with the DDS ops (NonMaxSuppression,
NonZero and RoiAlign) from TRT versions 10.0 to 10.7. TRT EP excludes
DDS ops from running on TRT by default, user can override default value
with empty string to include all ops.
ankitm3k pushed a commit to intel/onnxruntime that referenced this pull request Dec 11, 2024
…unning on TRT" (microsoft#22878)

- Revert microsoft#22681
- But still implicitly exclude DDS ops for TRT 10. Will later provide
better PR to add trt_op_types_to_exclude provider option.
ankitm3k pushed a commit to intel/onnxruntime that referenced this pull request Dec 11, 2024
…n TRT (microsoft#22681)

Add new provider option `trt_op_types_to_exclude`:
- User can provide op type list to be excluded from running on TRT
- e.g. `trt_op_types_to_exclude="MaxPool"`

There is a known performance issue with the DDS ops (NonMaxSuppression,
NonZero and RoiAlign) from TRT versions 10.0 to 10.7. TRT EP excludes
DDS ops from running on TRT by default, user can override default value
with empty string to include all ops.
ankitm3k pushed a commit to intel/onnxruntime that referenced this pull request Dec 11, 2024
…unning on TRT" (microsoft#22878)

- Revert microsoft#22681
- But still implicitly exclude DDS ops for TRT 10. Will later provide
better PR to add trt_op_types_to_exclude provider option.
ankitm3k pushed a commit to intel/onnxruntime that referenced this pull request Dec 11, 2024
…n TRT (microsoft#22681)

Add new provider option `trt_op_types_to_exclude`:
- User can provide op type list to be excluded from running on TRT
- e.g. `trt_op_types_to_exclude="MaxPool"`

There is a known performance issue with the DDS ops (NonMaxSuppression,
NonZero and RoiAlign) from TRT versions 10.0 to 10.7. TRT EP excludes
DDS ops from running on TRT by default, user can override default value
with empty string to include all ops.
ankitm3k pushed a commit to intel/onnxruntime that referenced this pull request Dec 11, 2024
…unning on TRT" (microsoft#22878)

- Revert microsoft#22681
- But still implicitly exclude DDS ops for TRT 10. Will later provide
better PR to add trt_op_types_to_exclude provider option.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants