allaccs

Worked yesterday for me, stopped working today. Workaround : git checkout e6a132a449d10489e65544e77f49190e2817b587 Use the commit from yesterday and it works, I guess an update broke the code.

QUESTION: What colour is the sky? ANSWER: It is commonly thought to be blue. QUESTION: Who first landed on the moon? ANSWER: A good question, this is known to be...

> Here are some of the reasons why it is a bad output for a language model: > > Incorrect answers: The model provides incorrect answers for some of the...

> Ya'll really got accepted as machine learning researchers? Smh > > First, lets stop comparing this to ChatGPT right now, less ya'll talking about your 6 gpu cluster running...

@timlacroix Thanks for the FAQ update. Changing the prompt formulation did give some much better outputs. I still find few shots prompting isn't giving very good results so I will...

No I mean actual fine-tuning, few-shot prompting seems to have it's limits. For Neo-x I have a 2go JSONL file with many documents, each document with a prompt/answer structure with...

3 months late mate, this issue was posted shortly after llama was out. I figured out (and others here for sure) that llama needed finetuning and RLHF to reach gpt...

I am having the same problem of repeated text, what is the current workaround?