[feat] add support for aten::reciprocal(int) #1308
Merged
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
narendasan
reviewed
Aug 25, 2022
| auto reciprocal_registration TORCHTRT_UNUSED = RegisterNodeConversionPatterns().pattern( | ||
| {"aten::reciprocal(Tensor self) -> Tensor", [](ConversionCtx* ctx, const torch::jit::Node* n, args& args) -> bool { | ||
| auto in = args[0].ITensorOrFreeze(ctx); | ||
| if (in->getType() == nvinfer1::DataType::kINT32) { |
There was a problem hiding this comment.
This might be a large change but would it make sense to just add this to the macro for other unary ops? @peri044 thoughts on what the repercussions would be?
There was a problem hiding this comment.
In this case the behavior matches pytorch. For other ops (ex. abs implemented element-wise above) this behavior would be incorrect. I have not checked any of the other ops.
There was a problem hiding this comment.
There are two ops which have some restrictions according to the doc but other unary ops must have floating point inputs. https://docs.nvidia.com/deeplearning/tensorrt/api/c_api/classnvinfer1_1_1_i_network_definition.html#a77831224c9a72ad02587a56ded93c672
Generally the input must have a floating-point type (or kINT8 as a quantized float), except for the following operations:
kSIGN accepts a floating-point or Int32 tensor.
kNOT requires a Bool tensor.
There was a problem hiding this comment.
We could add the above restrictions in the code in addition to what Michael added, to cover the cases in the doc completely.
There was a problem hiding this comment.
I was mostly talking from the pytorch perspective since seems like theres at least a few ops where int inputs are valid
narendasan
added a commit
that referenced
this pull request
Sep 9, 2022
* chore: additional options for perf_run tool Signed-off-by: dperi <dperi@nvidia.com> * feat: Add fx2trt backend and revamp current perf utility to accept CLI arguments Signed-off-by: Dheeraj Peri <peri.dheeraj@gmail.com> * chore: Refactor fx2trt functionality Signed-off-by: Dheeraj Peri <peri.dheeraj@gmail.com> * chore: Fix fp16 functionality for fx2trt backend Signed-off-by: Dheeraj Peri <peri.dheeraj@gmail.com> * chore: refactor Signed-off-by: Dheeraj Peri <peri.dheeraj@gmail.com> * chore: minor change Signed-off-by: Dheeraj Peri <peri.dheeraj@gmail.com> * refactor: Refactor perf_run and add internal benchmark scripts Signed-off-by: Dheeraj Peri <peri.dheeraj@gmail.com> * chore : minor refactor Signed-off-by: Dheeraj Peri <peri.dheeraj@gmail.com> * chore: Apply precommit tooling Signed-off-by: Dheeraj Peri <peri.dheeraj@gmail.com> * chore: Fix data loader issues and nox file paths Signed-off-by: Dheeraj Peri <peri.dheeraj@gmail.com> * chore: rebase and minor changes Signed-off-by: Dheeraj Peri <peri.dheeraj@gmail.com> * chore: Fix reporting to a file setting Signed-off-by: Dheeraj Peri <peri.dheeraj@gmail.com> * Update lower.py (#1324) * docs: [Automated] Regenerating documenation for e374eb1 Signed-off-by: Torch-TensorRT Github Bot <torch-tensorrt.github.bot@nvidia.com> * refactor: Refactor testing to use cosine similarity, remove redundancy models and restructuring Signed-off-by: Dheeraj Peri <peri.dheeraj@gmail.com> * chore: move to cosine similarity comparison Signed-off-by: Dheeraj Peri <peri.dheeraj@gmail.com> * refactor: Refactor nox file testing Signed-off-by: Dheeraj Peri <peri.dheeraj@gmail.com> * chore: add missing scripts Signed-off-by: Dheeraj Peri <peri.dheeraj@gmail.com> * chore: Linter fixes Signed-off-by: Dheeraj Peri <peri.dheeraj@gmail.com> * fix!: Fixed Windows compilation failures Signed-off-by: Anurag Dixit <a.dixit91@gmail.com> * chore: Minor fix Signed-off-by: Dheeraj Peri <peri.dheeraj@gmail.com> * chore: use rn18 instead of rn50 Signed-off-by: Dheeraj Peri <peri.dheeraj@gmail.com> * docs: [Automated] Regenerating documenation for a1a4786 Signed-off-by: Torch-TensorRT Github Bot <torch-tensorrt.github.bot@nvidia.com> * chore: Add cpp tests with cosine sim Signed-off-by: Dheeraj Peri <peri.dheeraj@gmail.com> * chore: linter fixes Signed-off-by: Dheeraj Peri <peri.dheeraj@gmail.com> * [feat] Add support for argmax and argmin (#1312) * [feat] Add support for argmax and argmin Adds support for aten::argmax and aten::argmin. Fixes # (issue) Please delete options that are not relevant and/or add your own. - Bug fix (non-breaking change which fixes an issue) - New feature (non-breaking change which adds functionality) - Breaking change (fix or feature that would cause existing functionality to not work as expected) - This change requires a documentation update - [ ] My code follows the style guidelines of this project (You can use the linters) - [ ] I have performed a self-review of my own code - [ ] I have commented my code, particularly in hard-to-understand areas and hacks - [ ] I have made corresponding changes to the documentation - [ ] I have added tests to verify my fix or my feature - [ ] New and existing unit tests pass locally with my changes - [ ] I have added the relevant labels to my PR in so that relevant reviewers are notified * move max.cpp tests to test_max.cpp no functional change * fix permissions on max.cpp * docs: [Automated] Regenerating documenation for 9db2852 Signed-off-by: Torch-TensorRT Github Bot <torch-tensorrt.github.bot@nvidia.com> * chore: Deepcopy other objects Signed-off-by: Dheeraj Peri <peri.dheeraj@gmail.com> * fix: Fix deepcopy issues of PTQ calibrators Signed-off-by: Dheeraj Peri <peri.dheeraj@gmail.com> * chore: linter fixes Signed-off-by: Dheeraj Peri <peri.dheeraj@gmail.com> * chore: Adding a guideline to build on Windows platform (#1337) * chore: Adding Windows build guideline Signed-off-by: Anurag Dixit <a.dixit91@gmail.com> * chore: Fix formatting Signed-off-by: Anurag Dixit <a.dixit91@gmail.com> Signed-off-by: Anurag Dixit <a.dixit91@gmail.com> * docs: [Automated] Regenerating documenation for 00a1f03 Signed-off-by: Torch-TensorRT Github Bot <torch-tensorrt.github.bot@nvidia.com> * chore: minor fixes Signed-off-by: Dheeraj Peri <peri.dheeraj@gmail.com> * chore: Linter fixes Signed-off-by: Dheeraj Peri <peri.dheeraj@gmail.com> * chore: Linter fixes Signed-off-by: Dheeraj Peri <peri.dheeraj@gmail.com> * docs: [Automated] Regenerating documenation for 1efe4b1 Signed-off-by: Torch-TensorRT Github Bot <torch-tensorrt.github.bot@nvidia.com> * docs: [Automated] Regenerating documenation for 10b9ecd Signed-off-by: Torch-TensorRT Github Bot <torch-tensorrt.github.bot@nvidia.com> * add support for aten::reciprocal(int) (#1308) * docs: [Automated] Regenerating documenation for 096fd41 Signed-off-by: Torch-TensorRT Github Bot <torch-tensorrt.github.bot@nvidia.com> Signed-off-by: dperi <dperi@nvidia.com> Signed-off-by: Dheeraj Peri <peri.dheeraj@gmail.com> Signed-off-by: Torch-TensorRT Github Bot <torch-tensorrt.github.bot@nvidia.com> Signed-off-by: Anurag Dixit <a.dixit91@gmail.com> Co-authored-by: dperi <dperi@nvidia.com> Co-authored-by: Dheeraj Peri <peri.dheeraj@gmail.com> Co-authored-by: Wei <wwei6@fb.com> Co-authored-by: Torch-TensorRT Github Bot <torch-tensorrt.github.bot@nvidia.com> Co-authored-by: Anurag Dixit <a.dixit91@gmail.com> Co-authored-by: Michael Feliz <104801882+mfeliz-cruise@users.noreply.github.com>
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.
Description
The unary layer does not support integer inputs to RECIP. Pytorch implicitly casts integer inputs to float for aten::reciprocal so we can add the same cast here to add support.
Fixes # (issue)
Type of change
Please delete options that are not relevant and/or add your own.
Checklist: