vicgalle

/

gpt2-open-instruct-v1

like 31

Hi @suriya7 . Sadly, I lost the fine-tune code for this model, as I trained it over a year ago.
Currently, one of the easiest ways would be to use the SFTTrainer from the trl library: https://huggingface.co/docs/trl/sft_trainer

yah @vicgalle I have trained GPT-2 medium, but my loss can't converge to less than 1.0. That's why I asked for the code to see what techniques you have implemented.🙃