(모델 요약) LLaMa3 (paper)

  • param size: 8B, 70B
  • context length: 8k
  • pretraining data: 15T+ tokens of data from publicly available sources
  • fine-tuning data: publicly available instruction datasets, as well as over 10M human-annotated examples
  • Python 으로 경량화 모델 로드: ollama