kto训练要求response大于1(feedback_dataset)函数

[tcxia]

Reminder

• [X] I have read the README and searched the existing issues.

System Info

Reproduction

NPROC_PER_NODE=7 NNODES=1 RANK=0 MASTER_ADDR=127.0.0.1 MASTER_PORT=29504

CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6 torchrun \ --nproc_per_node $NPROC_PER_NODE \ --nnodes $NNODES \ --node_rank $RANK \ --master_addr $MASTER_ADDR \ --master_port $MASTER_PORT \ src/train.py examples/aigc_train/llama3/llama3_lora_kto.yaml | tee train_aigc_kto.log

Expected behavior

No response

Others

No response

[hiyouga]

数据集格式不对

[tcxia]

是按照这个来配置的，但是依然报错

[hiyouga]

参考 https://github.com/hiyouga/LLaMA-Factory/blob/main/data/kto_en_demo.json

[tcxia]

这个属于sharegpt格式？

[xuguozhi]

这个属于sharegpt格式？

是不是lable没有用的bool，而是str了