QLinearPackWeightInt8 Class — pytorch Architecture

Source Code

aten/src/ATen/native/ao_sparse/quantized/cpu/qlinear_prepack.cpp lines 212–238

class QLinearPackWeightInt8 final {
 public:
  static c10::intrusive_ptr<LinearPackedParamsBase> run(
      const at::Tensor& weight,
      const std::optional<at::Tensor>& bias,
      const int64_t out_features_block_size,
      const int64_t in_features_block_size) {
    auto& ctx = at::globalContext();

#ifdef USE_FBGEMM
    if (ctx.qEngine() == at::QEngine::FBGEMM) {
      return PackedLinearWeight::prepack(
          weight, bias, out_features_block_size, in_features_block_size);
    }
#endif
#ifdef USE_PYTORCH_QNNPACK
    if (ctx.qEngine() == at::QEngine::QNNPACK) {
      return PackedLinearWeightQnnp::prepack(
          weight, bias, out_features_block_size, in_features_block_size);
    }
#endif
    TORCH_CHECK(
        false,
        "Didn't find engine for operation ao::sparse::qlinear_prepack ",
        toString(ctx.qEngine()));
  }
};