qgyd2021/sft_llama2_stack_exchange

I followed this script to train this model.

instead of the official meta-llama/Llama-2-7b-hf model, I used this repo NousResearch/Llama-2-7b-hf.

The model trained on lvwerra/stack-exchange-paired dataset.

seq_length: 1024

steps: 1600