Finetuning with custom data and different LLM

[imerad]

Hello,

Thank you for making your code public. I would like to finetune the machine translation model on custom data for a little known language. However, I don't understand the format in which the training data should be provided. Could you please provide more details about what the .pt files should contain exactly ?

Also, is it possible to use the more recent Llama 3.1 8b model ? How ?

Thanks