Add QInt8, QUInt8, QInt32 quantized scalar types with full quantizati…#1531
Merged
alinpahontu2912 merged 2 commits intodotnet:mainfrom Feb 24, 2026
Merged
Add QInt8, QUInt8, QInt32 quantized scalar types with full quantizati…#1531alinpahontu2912 merged 2 commits intodotnet:mainfrom
alinpahontu2912 merged 2 commits intodotnet:mainfrom
Conversation
Contributor
There was a problem hiding this comment.
Pull request overview
This pull request adds comprehensive quantization support to TorchSharp, enabling the use of quantized scalar types (QInt8, QUInt8, QInt32) and associated quantization operations. The changes align TorchSharp's quantization API with PyTorch's native quantization capabilities.
Changes:
- Adds three quantized scalar types (QInt8, QUInt8, QInt32) to the ScalarType enum with correct ordinal values matching PyTorch
- Implements full quantization API including per-tensor and per-channel quantization/dequantization operations with corresponding native bindings
- Adds 13 comprehensive unit tests covering all new functionality including edge cases and error handling
Reviewed changes
Copilot reviewed 7 out of 7 changed files in this pull request and generated no comments.
Show a summary per file
| File | Description |
|---|---|
| test/TorchSharpTest/TestTorchSharp.cs | Adds 13 comprehensive unit tests for quantization operations covering enum values, dtype aliases, type checking, quantization/dequantization, and per-channel operations |
| src/TorchSharp/Tensor/torch.PointwiseOps.cs | Adds public API methods for quantize_per_tensor, quantize_per_channel, and dequantize with proper validation and documentation |
| src/TorchSharp/Tensor/TensorExtensionMethods.cs | Adds IsQuantized extension method for ScalarType to support internal type checking |
| src/TorchSharp/Tensor/Tensor.cs | Adds quantized scalar type enum values, dtype aliases, is_quantized checking, and Tensor instance methods for quantization operations (dequantize, q_scale, q_zero_point, int_repr, q_per_channel_*) |
| src/TorchSharp/PInvoke/LibTorchSharp.THSTensor.cs | Adds P/Invoke declarations for 9 new native quantization functions with proper marshalling signatures |
| src/Native/LibTorchSharp/THSTensor.h | Adds C++ function declarations for quantization operations following existing naming conventions |
| src/Native/LibTorchSharp/THSTensor.cpp | Implements native quantization bindings using CATCH_TENSOR and CATCH_RETURN macros for consistent error handling |
💡 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 no new comments.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
alinpahontu2912
force-pushed
the
qint_type
branch
from
a103b80 to
54b5302
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 2 comments.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
…on support - Uncomment and fix ScalarType enum entries (QInt8=12, QUInt8=13, QInt32=14) - Fix incorrect QUInt32 name to QInt32 to match PyTorch - Add torch.qint8, torch.quint8, torch.qint32 dtype aliases - Add torch.is_quantized() and Tensor.is_quantized() methods - Add IsQuantized() extension method on ScalarType Native quantization bindings: - Add THSTensor_quantize_per_tensor, THSTensor_quantize_per_channel (C++ and P/Invoke) - Add THSTensor_dequantize (C++ and P/Invoke) - Add THSTensor_q_scale, THSTensor_q_zero_point (C++ and P/Invoke) - Add THSTensor_int_repr (C++ and P/Invoke) - Add THSTensor_q_per_channel_scales, THSTensor_q_per_channel_zero_points, THSTensor_q_per_channel_axis (C++ and P/Invoke) Managed API: - Add torch.quantize_per_tensor() and torch.quantize_per_channel() static methods - Add torch.dequantize() static method - Add Tensor.dequantize(), Tensor.q_scale(), Tensor.q_zero_point() instance methods - Add Tensor.int_repr() instance method - Add Tensor.q_per_channel_scales(), Tensor.q_per_channel_zero_points(), Tensor.q_per_channel_axis() instance methods Unit tests for all new functionality (13 tests)
…e leak - Add QInt8 (1), QUInt8 (1), QInt32 (4) to ElementSize() so it no longer throws NotImplementedException for quantized types - Split chained tensor().reshape() in QuantizePerChannel test to properly dispose the intermediate 1D tensor via 'using var' Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
alinpahontu2912
force-pushed
the
qint_type
branch
from
b2c45d3 to
5f483cd
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 #141
Native quantization bindings:
Managed API:
Unit tests for all new functionality (13 tests)