Changing channels of FNO in Modulus

Hello,

we are trying to train a FNO model with the input has 3 channels, but we are facing problems when the model is doing convolution. It will give out an error “RuntimeError: Given groups=1, weight of size [32, 3, 1, 1], expected input[32, 5, 100, 200] to have 3 channels, but got 5 channels instead”. When we switch to input input with 2 channels, we got a similar error “RuntimeError: Given groups=1, weight of size [32, 3, 1, 1], expected input[32, 4, 100, 200] to have 3 channels, but got 4 channels instead”.
We set the input channel variable in the training script, but nothing changes.

Does anyone know if we can input data with 3 channels for FNO in Modulus? I am pretty sure that we can do so in the original FNO model, but we are not sure if we need to change something in the Modulus?

Thanks