Conv2dtranspose in tvm relay

Hello, When I tried to import tensorflow graph to relay frontend, the following error comes out. The tensorflow graph has conv2d_transpose, which tensor shape is (2,2,512,1024). In Tensorflow/Keras conv2d_transpose, the tensor shape is (H,W, output_channel,input_channel). This is different than general conv2d tensor shape, which is (H,W, input_channel, output_channel). The error is due to the mismatch of the tensor shape. Is there any plan to support conv2d_transpose in TVM relay?

%34 = transpose(%33, axes=[0, 3, 1, 2]); %35 = transpose(%Variable_18, axes=[3, 2, 0, 1]); %36 = nn.conv2d_transpose(%34, %35, channels=512, kernel_size=[2, 2], strides=[2, 2], padding=[0, 0, 0, 0], kernel_layout=“IOHW”) in particular dimension 0 conflicts 512 does not match 1024dimension 1 conflicts 1024 does not match 512; unable to unify: Tensor[(512, 1024, 2, 2), float32] and Tensor[(1024, 512, 2, 2), float32]; ; %37 = transpose(%36, axes=[0, 2, 3, 1]); %38 = add(%37, %Variable_41); %39 = (%26, %38); concatenate(%39, axis=3) }