Packed weight matrix for DNNLOWP Int8Conv operator. More...

Inheritance diagram for caffe2::Int8ConvDNNLowPPackedWeightBlob:

Data Fields
std::shared_ptr< fbgemm::Packed3x3ConvMatrix >	W_depthwise_3x3

std::shared_ptr< fbgemm::Packed3x3x3ConvMatrix >	W_depthwise_3x3x3

std::shared_ptr< fbgemm::PackWeightMatrixForGConv< std::int8_t > >	W_gconv

Data Fields inherited from caffe2::Int8FCDNNLowPPackedWeightBlob
std::vector< dnnlowp::TensorQuantizationParams >	qparams

std::shared_ptr< std::vector< std::int32_t > >	column_offsets

Tensor	original_tensor {CPU}

std::shared_ptr< std::vector< std::int32_t > >	bias

std::shared_ptr< fbgemm::PackBMatrix< std::int8_t > >	W

std::shared_ptr< fbgemm::PackBMatrix< std::int8_t, std::int16_t > >	W_acc16

std::shared_ptr< fbgemm::CompressedSparseColumn >	W_outlier

int	nbits_in_non_outlier

Detailed Description

Packed weight matrix for DNNLOWP Int8Conv operator.

Definition at line 38 of file fbgemm_pack_blob.h.

The documentation for this struct was generated from the following file:

Data Fields