blackkker opened a new pull request, #12467: URL: https://github.com/apache/tvm/pull/12467
According to the pytorch doc [torch.nn.MaxPool3d](https://pytorch.org/docs/stable/generated/torch.nn.MaxPool3d.html?highlight=pool3d#torch.nn.MaxPool3d). The shape of input can be `(N,C,D,H,W) or (C,D,H,W)`. While the MaxPool3DAttrs thinks the dimension should be 5.https://github.com/apache/tvm/blob/bcc7cde95c1e84b85f18c07110489350865b8cfe/include/tvm/relay/attrs/nn.h#L985-L996 which could leads to out of bound error `Check failed: (0 <= i && i < p->size_) is false: IndexError: indexing 4 on an array of size 4` when `ii == 4 and dshape == 4`. https://github.com/apache/tvm/blob/bcc7cde95c1e84b85f18c07110489350865b8cfe/src/relay/op/nn/pooling.cc#L1211-L1216 Therefore, I add an expand_dims operation on the front end to avoid this situation. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
