Fine-tune Paligemma model for segmentation task is failing

Hello:
I want to use Paligemma to segment water in satellite images. However, I haven't been able to find any documentation on how to scale the points inside my mask to be able to use them in Paligemma.
So far I have done the following:

Variables:
(x, y): is 1 point of the 8 needed.
w, h = width, height of the bounding box

I do the following transformation of x and y:

x_paligemma = xw / 64
y_paligemma = yh / 64

When I try to fine-tune the Paligemma model on this transformation, I obtain empty predictions.

Could you please tell me how can I prepare the segmentation points to have results in my model?

Thank you so much for your response!

The issue names no repository files, tests, or entry points. Start by locating the PaliGemma segmentation fine-tuning examples and documentation, then compare the reported point transformation with the expected input format; done means obtaining non-empty segmentation predictions for the stated satellite-image task.

Written by the indexing model from the issue text.