费政聪

Actually, toy_data.json is the first 100 items in the train.json, which is used to debug code quickly. You can use train.json instead, off course, we will fix it later :)

Hi, i have tried to fine-tune LLaMA model for better Chinese text generation. Please refer to: https://github.com/feizc/MLE-LLaMA

Hi, i have tried for instruct tuning on LLaMa, and provides scripts and alignment datasets : ) . You can reference for: https://github.com/feizc/MLE-LLaMA

> @feizc how are you approaching the problem of generating starting from a length that is less than the prefix? Actually, I use a fixed length of conditional context, i.e.,...