Hi All - I am new to Apache TVM and tinkering around. I took an NVIDIA T4 machine on AWS. I replicated the ResNet50 example. Native PyTorch is giving 87ms while TVM tuned is giving 101ms.
Then I did the same on my local computer. AMD Ryzen 16 core CPU. There the native Pytorch is around 102ms but the TVM tuned is around 40ms.
Why this massive difference?