Text Generation
Transformers
Safetensors
English
llama
nvidia
llama3.1
conversational
text-generation-inference

mistakes, inserts Chinese characters and flood

#64
by Markobes - opened

Hi,
I tested on HF-chat the model on the translation of a fragment of subtitles.
The model makes logical mistakes in the text and inserts Chinese characters* in response.
And this: ". the, the self is the the. the the. **nt the. the. the the.l.ily the.f the. the the, self. F. f the that's perspective. this. **"
*It looks like you've outsourced model development to China. Don't be surprised now by the fall of your shares on the stock exchange.

Regards

Also getting this error multiple times with quite a few prompts, but the moment I test it with just a 'Hello?'as the prompt, it acts normal as if nothing wierd was responded with.

FIX ASAP!

Sign up or log in to comment