QLinearInt8FusedQDQ Class — pytorch Architecture

Source Code

aten/src/ATen/native/quantized/cpu/qlinear.cpp lines 1561–1576

class QLinearInt8FusedQDQ final {
 public:
  static at::Tensor run(
      at::Tensor input,
      double input_scale,
      int64_t input_zero_point,
      const c10::intrusive_ptr<LinearPackedParamsBase>& packed_weight) {
    if (ReluFused) {
      return packed_weight->apply_with_input_q_dq_qweight_dq_relu_output_fp32(
          std::move(input), input_scale, input_zero_point);
    } else {
      return packed_weight->apply_with_input_q_dq_qweight_dq_output_fp32(
          std::move(input), input_scale, input_zero_point);
    }
  }
};