vllm

how many languages supported?

#16
by xingwang1234 - opened

This model can process images. If the image contains text, how many languages can the model perform OCR on?

I think it only understand languages written in latin or cyrillic alphabets. (I bet greek works too)

Other types of writing systems are not recognized In my testing. At least Arabic, Chinese and Japanese characters are causing very bad hallucinations.
I haven't tried Korean Hanguls yet, maybe it could works since it's technically an alphabet too, but I wouldn't bet on it. Korean is not working either.
I haven't tried Indian or Bengali characters either.

Just got to test with Greek and Hindi. It understands the former, but not the later.

Sign up or log in to comment