Closed K-H-Ismail closed 3 years ago
Hello,
I have noticed that in https://github.com/chengdazhi/Deformable-Convolution-V2-PyTorch/blob/47e4745a51d1a35af51f2f6960ad0cf9c7a6afed/src/cuda/deform_conv_cuda.cu#L76
we create an empty "at" tensor named "output", but then at the end of the forward function https://github.com/chengdazhi/Deformable-Convolution-V2-PyTorch/blob/47e4745a51d1a35af51f2f6960ad0cf9c7a6afed/src/cuda/deform_conv_cuda.cu#L117
we return a view of it without filling it properly with the result. Am I wrong ? Could you clarify this please ?
Hello,
I have noticed that in https://github.com/chengdazhi/Deformable-Convolution-V2-PyTorch/blob/47e4745a51d1a35af51f2f6960ad0cf9c7a6afed/src/cuda/deform_conv_cuda.cu#L76
we create an empty "at" tensor named "output", but then at the end of the forward function https://github.com/chengdazhi/Deformable-Convolution-V2-PyTorch/blob/47e4745a51d1a35af51f2f6960ad0cf9c7a6afed/src/cuda/deform_conv_cuda.cu#L117
we return a view of it without filling it properly with the result. Am I wrong ? Could you clarify this please ?