LLaMA2 Fine-tuning for GSM-8K
First, create a conda environment llama2
, and then install pytorch library.
Second, install environment by pip install deepspeed
Third, install extra library by pip install -r requirement.txt
Fourth, enter the file named step1_supervised_finetuning
Fifth, exchange all the path to your local path, and execute bash ./training_scripts/train.sh
Sixth, evalute the trained llama2 on the test dataset by bash ./evaluation_scripts/eval.sh