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

[dandanxiao]

Hello, thank you very much for your open source code. When reproducing the experimental results, I found that the bleu scores of some data sets are very different from those in the paper. For example, in the fleurs data set, en->zh, its bleu score is only 6.39, and en-ja, its bleu score is only 1.09. I would like to ask if this is a problem with my training or the data set you gave is wrong? If possible, please send the trained model weights and the code for processing the data set to the mailbox 1945921138@qq.com, thank you very much!