Missing keys in loading pretrianed weights

[GZWQ]

Hi,

I received errors when loading the following models:

--source="openclip" --from_pretrained="xlm-roberta-base-ViT-B-32@laion5b_s13b_b90k"
--source="nllb-openclip" --from_pretrained="nllb-clip-base-siglip@v1"

The errors are:
RuntimeError: Error(s) in loading state_dict for CustomTextCLIP:
Missing key(s) in state_dict: "text.transformer.embeddings.position_ids".

Do you have any ideas of why the loading failed?

[gregor-ge]

Your version of open-clip-torch might be too old. Try updating it.

I can load the xlm-roberta-base-ViT-B-32@laion5b_s13b_b90k model without issues in version 2.27.0 and 2.30.0 (the newest one).