gogpt-560m / README.md
leaderboard-pr-bot's picture
Adding Evaluation Results
dc1416b
|
raw
history blame
1.79 kB
metadata
license: apache-2.0
datasets:
  - BelleGroup/train_2M_CN
  - BelleGroup/train_3.5M_CN
  - BelleGroup/train_1M_CN
  - BelleGroup/train_0.5M_CN
  - BelleGroup/school_math_0.25M
language:
  - zh

GoGPT

基于中文指令数据微调BLOOM img.png

训练第一轮足够了,后续第二轮和第三轮提升不大

  • 🚀多样性指令数据
  • 🚀筛选高质量中文数据
模型名字 参数量 模型地址
gogpt-560m 5.6亿参数 🤗golaxy/gogpt-560m
gogpt-3b 30亿参数 🤗golaxy/gogpt-3b

测试效果

img.png img.png img.png img.png img.png img.png

TODO

  • 进行RLFH训练
  • 后续加入中英平行语料

感谢

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric Value
Avg. 26.3
ARC (25-shot) 26.37
HellaSwag (10-shot) 31.86
MMLU (5-shot) 25.29
TruthfulQA (0-shot) 43.12
Winogrande (5-shot) 50.75
GSM8K (5-shot) 0.0
DROP (3-shot) 6.7