What fixed it for me:
In the feature_extraction_utils.py of the transformers library in line 141 changed
return torch.tensor(value)
to
return torch.tensor(value, dtype=torch.float16)
to fix the following error:
PaliGemma\.venv\Lib\site-packages\transformers\feature_extraction_utils.py", line 142, in as_tensor
return torch.tensor(value)
^^^^^^^^^^^^^^^^^^^
RuntimeError: Could not infer dtype of numpy.float32
Not a permanent fix. Works for the paligemma model card examples though.