projects / platform / upstream / pytorch.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 63c90ec) | patch
[quant][fx] Add support for dynamic linear + relu fusion (INT8) (#63799)
authorSupriya Rao <supriyar@fb.com>
Fri, 27 Aug 2021 04:05:56 +0000 (21:05 -0700)
committerFacebook GitHub Bot <facebook-github-bot@users.noreply.github.com>
Fri, 27 Aug 2021 04:10:46 +0000 (21:10 -0700)
commitc7027f19efbb2f7b274c9e5fc0e87fe4b084e6ae
tree030160e2cb8fe22b087efd723d28e3a92562bff4tree | snapshot
parent63c90ec3bf6c9445a36199f65e0523a5e6532b0dcommit | diff
[quant][fx] Add support for dynamic linear + relu fusion (INT8) (#63799)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/63799

Add a new module that can be used for module swap with the nni.LinearReLU module in convert function.
Supports INT8 currently (since FP16 op doesn't have relu fusion yet).

Fixes #55393

Test Plan:
python test/test_quantization.py test_dynamic_fusion

Imported from OSS

Reviewed By: heitorschueroff

Differential Revision: D30502812

fbshipit-source-id: 3668e4f001a0626d469e17ac323acf582ee28a51
test/quantization/eager/test_quantize_eager_ptq.py diff | blob | history
test/quantization/fx/test_quantize_fx.py diff | blob | history
torch/nn/intrinsic/quantized/dynamic/__init__.py [new file with mode: 0644] blob
torch/nn/intrinsic/quantized/dynamic/modules/__init__.py [new file with mode: 0644] blob
torch/nn/intrinsic/quantized/dynamic/modules/linear_relu.py [new file with mode: 0644] blob
torch/nn/quantized/dynamic/modules/linear.py diff | blob | history
torch/quantization/fx/quantization_patterns.py diff | blob | history
torch/quantization/ns/mappings.py diff | blob | history
torch/quantization/quantization_mappings.py diff | blob | history
torch/testing/_internal/common_quantization.py diff | blob | history
Domain: Machine Learning / ML Framework;