The ONNX documentation you wrote describes the reshaping that is done by their softmax implementation: an input tensor is always reshaped to ...
確定! 回上一頁