Inference

[TheoR14]

Hello, I finetuned the model (multi-temporal-crop-classification) with my data (training and validation chips with 224x224x pixels) and when I try the inference on other data (with 224x166 pixels) it doesn't work. I saw in the hls-foundation Repo that for the inference can be used on new data in GeoTIFF format and that data can be of any shape (e.g. height and width) as long as it follows the bands/channels of the original dataset.
I my case, the only things that chages is the image size, the bands are the same.

Do you have an idea why it doensn't work ?

Thanks
TheoR14