src/convolutional_layer.c
@@ -148,7 +148,9 @@ cudnnDataType_t data_type = CUDNN_DATA_FLOAT; #endif // Tensor Core uses CUDNN_TENSOR_OP_MATH instead of CUDNN_DEFAULT_MATH #if(CUDNN_MAJOR >= 7) cudnnSetConvolutionMathType(l->convDesc, CUDNN_TENSOR_OP_MATH); #endif // INT8_CONFIG, INT8_EXT_CONFIG, INT8x4_CONFIG and INT8x4_EXT_CONFIG are only supported // on architectures with DP4A support (compute capability 6.1 and later).
