Swiching the deggre of the data causes incorrect output

[ShileiCao]

The model seems to work well with the ERA5 0.25° data with pre-trained model following the official instructions. When I interpolate the 0.25° input data (721, 1440) into 1.5° input data (121, 240) with the interpolate functions provided in batch.py, the model prediction become：

Is this normal? Or we need to finetune model when we change the degree