Add approximate parameter to GELU activation function#1548
Merged
alinpahontu2912 merged 4 commits intoMay 4, 2026
Merged
Conversation
Contributor
There was a problem hiding this comment.
Pull request overview
Adds support for PyTorch’s approximate mode to GELU (notably "tanh"), threading the option through the native (C++), P/Invoke, Tensor, functional, and module APIs, and adding a regression test.
Changes:
- Introduces
Modules.GELU.Approximate(none/tanh) and plumbs it throughnn.GELUandnn.functional.gelu. - Extends Tensor
gelu/gelu_to accept an approximation mode and updates the corresponding native/PInvoke signatures. - Adds a unit test validating the tanh approximation path and that it differs from the exact mode.
Reviewed changes
Copilot reviewed 6 out of 6 changed files in this pull request and generated 2 comments.
Show a summary per file
| File | Description |
|---|---|
| test/TorchSharpTest/NN.cs | Adds a test covering GELU tanh approximation behavior. |
| src/TorchSharp/Tensor/Tensor.cs | Adds gelu/gelu_ overloads that pass approximation through to native. |
| src/TorchSharp/PInvoke/LibTorchSharp.THSTensor.cs | Updates P/Invoke signatures to accept the approximation string. |
| src/TorchSharp/NN/Activation/GELU.cs | Adds approximation enum + overloads in module factory and functional API. |
| src/Native/LibTorchSharp/THSTensor.h | Updates native exports for GELU to accept an approximation parameter. |
| src/Native/LibTorchSharp/THSTensor.cpp | Passes approximation through to torch::gelu / torch::gelu_. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
alinpahontu2912
force-pushed
the
feature/gelu-approximate-parameter
branch
from
0079b80 to
a098ea4
Compare
Contributor
There was a problem hiding this comment.
Pull request overview
Copilot reviewed 7 out of 7 changed files in this pull request and generated 4 comments.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Contributor
There was a problem hiding this comment.
Pull request overview
Copilot reviewed 7 out of 7 changed files in this pull request and generated 3 comments.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
alinpahontu2912
force-pushed
the
feature/gelu-approximate-parameter
branch
from
cf93772 to
3ae09b0
Compare
Contributor
There was a problem hiding this comment.
Pull request overview
Copilot reviewed 7 out of 7 changed files in this pull request and generated 1 comment.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
cincuranet
approved these changes
Apr 10, 2026
cincuranet
approved these changes
Apr 28, 2026
Add support for the 'approximate' parameter in GELU, matching PyTorch's torch.nn.GELU(approximate='tanh') functionality. Changes: - Add GELU.Approximate enum with 'none' and 'tanh' values - Thread approximate parameter through all layers: native C++, PInvoke, Tensor methods, functional API, and module factory - Add new overloads (no breaking changes to existing API) - Add test for tanh approximation mode Fixes dotnet#1368 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
- Move Approximate enum from GELU module class to neutral TorchSharp namespace as GELUApproximate, removing Tensor/functional layer dependency on Modules layer - Add CharSet, BestFitMapping, ThrowOnUnmappableChar attributes to THSTensor_gelu/gelu_ DllImport declarations to match existing LPStr-based imports pattern - Update all references in Tensor.cs, GELU.cs, and tests Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
- Keep original THSTensor_gelu/gelu_ exports unchanged for ABI compatibility - Add new THSTensor_gelu_with_approximate/gelu_with_approximate_ exports that accept the approximate string parameter - Add null guard in native code, treating null as 'none' - Update P/Invoke declarations and managed callers accordingly Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Move the standalone GELUApproximate enum into the GELU class as a nested Approximate enum, making the public API surface GELU.Approximate instead of GELUApproximate. This matches the intended API design and scopes the enum to its semantically relevant module. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
alinpahontu2912
force-pushed
the
feature/gelu-approximate-parameter
branch
from
6d3eac0 to
64db9b2
Compare
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
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.
Fixes #1368
Add support for the 'approximate' parameter in GELU, matching PyTorch's torch.nn.GELU(approximate='tanh') functionality.
Changes: