Hello, why is the F1 value of the test 0?

[hcztxdy]

[Inkyl]

Thanks for reaching out! On our side, the training metrics are not zero, so the model is learning as expected. However, based on your description, it seems that your loss values are abnormal (e.g., not changing). This could indicate an issue with the training setup.

Could you confirm if you are using the ConvNeXt and Segformer pretrained weights? Pretrained weights are crucial for proper initialization, and skipping them might lead to unexpected behavior.

If you have already used the pretrained weights and are still facing problems, please share more details (e.g., your training setup, logs, or error messages). This will help us better understand the issue and provide further assistance.

Looking forward to your reply!