pytorch
e62bf892 - Renaming variables from dX to dY in Learnable Fake Quantize kernels for Better Clarity (#42032)

Commit
4 years ago
Renaming variables from dX to dY in Learnable Fake Quantize kernels for Better Clarity (#42032) Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/42032 In this diff, the arguments `dX` within the C++ kernels are named as `dY` for clarity and avoid confusion since it doesn't represent the gradient with respect to the input. Test Plan: To test all related fake quantize kernel operators, on a devvm, run the command: `buck test //caffe2/test:quantization -- learnable` Reviewed By: z-a-f, jerryzh168 Differential Revision: D22735429 fbshipit-source-id: 9d6d967f08b98a720eca39a4d2280ca8109dcdd6
Author
Paul Shao
Parents
  • aten/src/ATen/native/quantized
    • cpu/kernels
      • File
        QuantizedOpKernels.cpp
    • cuda
      • File
        fake_quantize_core.cu
    • File
      fake_quant_per_channel_affine.cpp