Windows7 64 bit
model.qconfig = torch.quantization.default_qconfig
Exception has occurred: RuntimeError
Didn’t find engine for operation quantized::conv2d_prepack NoQEngine
Could you check if the
USE_FBGEMM macro is turned on in your build environment? I believe it should be supported.
cc @dskhudia who might be able to add more info.
We use VS 14.11 to build binaries for CUDA 9.2, so there is no FBGEMM support. If you need FBGEMM, then please use the binaries with other CUDA versions instead.