hiyouga / LLaMA-Factory

Unify Efficient Fine-Tuning of 100+ LLMs
Apache License 2.0
25.52k stars 3.16k forks source link

kto训练完如何预测 #4600

Closed tcxia closed 5 days ago

tcxia commented 5 days ago

Reminder

System Info

如题,如何利用kto结果对数据进行评估

Reproduction

CUDA_VISIBLE_DEVICES=0 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_predict.yaml | tee infer_aigc_kto_llama3.log

Expected behavior

No response

Others

No response

hiyouga commented 5 days ago

用 examples 里面的批量预测

tcxia commented 5 days ago

image KTO的运行脚本里面,好像没有do_predict的选项?

hiyouga commented 5 days ago

用 sft 的