I built a ResNet8 model using AdaPT_Conv2d and AdaPT_Linear layers. During training, the model's accuracy did not improve, so I reviewed the backward function implementations in the files cpu-kernels/axx_linear.cpp and cpu-kernels/axx_conv2d.cpp. It seems that the backward function in axx_linear.cpp is incomplete, and it is entirely undefined in axx_conv2d.cpp. Am I missing something, or is backward propagation not functioning for these layers? Thank you for any response.
I built a ResNet8 model using AdaPT_Conv2d and AdaPT_Linear layers. During training, the model's accuracy did not improve, so I reviewed the backward function implementations in the files cpu-kernels/axx_linear.cpp and cpu-kernels/axx_conv2d.cpp. It seems that the backward function in axx_linear.cpp is incomplete, and it is entirely undefined in axx_conv2d.cpp. Am I missing something, or is backward propagation not functioning for these layers? Thank you for any response.