etri-xainlp/llama2-13b-lima-sft-dpo

Model Details

Model Developers ETRI xainlp team

Input text only.

Output text only.

Model Architecture

Base Model meta-llama/Llama-13b-hf

Training Dataset

• fully sft: 650k instruction-following set

• lima sft: 280k instruction-following set

• dpo+lora: 90k user preference set

• We use A100 GPU 80GB * 7, when training.