Closed pourmand1376 closed 1 year ago
Hi @pourmand1376, sorry for a slighly off-topic question: could you please share any details on how your friend managed to fine-tune LLaMA on text-only dataset, without instructions? I'm interested in doing the same thing with Belarusian Wikipedia, but so far I've only seen tutorials on how to instruct-tune LLaMA, and Wikipedia articles as such don't contain clearly delimited prompts and responses. Could you please briefly describe the approach?
Thanks in advance for any comments.
Currently, the Open-assistant model doesn't support Farsi. This is a text-only dataset to learn Farsi (Persian).
One of my friends fine-tuned LLaMa on this dataset and It could understand Farsi grammar and word usage very well. If the Open-assistant team wants to add support to Farsi, this should be the first step.
I have transformed the dataset into the standard that has been mentioned here and uploaded it to my huggingface account.
2974