tensorbuffer
tensorbuffer
From this: https://docs.nvidia.com/deeplearning/performance/dl-performance-convolutional/index.html Convolution of an NCHW input tensor with a KCRS weight tensor, producing a NKPQ output From the onnx model, we have NCHW = 1,1,327,80, KCRS = 128,...
ah, this works! What was the issue, you think the input is not in NCHW format? Earlier I also tried to set the input shape to 128 instead of 327,...
Thanks very much for the detailed explanation, I will read it through and check the code. Thanks! This helped us a lot!
Yes I can repro with your steps. I see you added TODO label, so I guess this will be fixed?