llamaindex

/

vdr-2b-multi-v1

like 117

Follow

LlamaIndex 167

This is because of the new transformer version that was released recently.
https://pypi.org/project/transformers/#history
If you specify version 4.49 or lower it'll work, I don't know how to fix this in your specific setup but there must be an issue on llamaindex side where they don't specifiy the version of the transformers package they are using.

Fixed here https://huggingface.co/llamaindex/vdr-2b-multi-v1/commit/4d9a12ebde4251fd9a9ac45b463ba8d89e73a153

Hi @cheesyFishes and llamaindex team, please can you fix the same issue in llamaindex/vdr-2b-v1 ?

Thanks.

in transformers ≥ 4.50 image_processing_utils.get_size_dict() allows only these key-sets:

•	{"height","width"}
•	{"shortest_edge"}
•	{"shortest_edge","longest_edge"}
•	{"longest_edge"}
•	{"max_height","max_width"}

So the model is still getting ValueError.
Current preprocessor_config.json has four keys (max_pixels, longest_edge, min_pixels, shortest_edge), so it still fails the above check.