Hangs on input longer than 8192 characters

[simonw]

I tried running this in the https://github.com/simonw/hmb-map folder (to get all the node_models READMEs):

llm embed-multi jina-readmes \
      -m jina-embeddings-v2-small-en \
      --files . '**/README.md' --store \
      --database test.db

I got this:

Embedding [####################################] 100%Token indices sequence length is longer than the specified maximum sequence length for this model (8367 > 8192). Running this sequence through the model will result in indexing errors
/Users/simon/.cache/huggingface/modules/transformers_modules/jinaai/jina-bert-implementation/a9db86227f71a0bd7bc05e5dda0359f1e09abb0f/modeling_bert.py:774: UserWarning: Increasing alibi size from 8192 to 8367.
warnings.warn(

The process seemed to hang - I had to Ctrl+Z and then kill %1 to exit it.

[simonw]

Truncating to 8192 characters before passing it to the model seemed to fix it.

[simonw]

I couldn't get the new test to fail but I did see it fail and then work testing manually on the CLI.