数据集:
lucasmccabe-lmi/FLAN_CoT_alpaca_style
预印本库:
arxiv:2210.11416We provide a dataset representing the 9 chain-of-thought (reasoning) fine-tuning tasks from FLAN . Minor formatting has been applied:
Numbers:
Prompts: 74771
Tokens: 9016176 using the EleutherAI/gpt-neox-20b tokenizer (counting instruction+input+output)