Ade20k dataset val/test problem

[gavin-gqzhang]

When I used segformer to directly use the Ade20k dataset for training and testing, it was shown that there was a difference in size. I tried to test_ Add resize to pipeline, but it can't be solved

[gavin-gqzhang]

Error reporting content: RuntineError: stack expects each tensor to be equal size, but got [3，512，512] at entry 8 and [3,704,512]at entry 1