-
我要微调reranker,你的介绍里也说了可以调bge-reranker-large,train_reranker.sh文件里根本没有这个接口啊?我用的第一个,提示bge-reranker-large没有bertlayer类 #model_bert,fsdp(ddp) CUDA_VISIBLE_DEVICES="1" nohup accelerate launch --config_file ../../../config/default_fsdp.yaml train_reranker.py \#--model_name_or_path "/data/liuxiang/58/10_llm/6_比赛/bge-reranker-large" >./logs/t2ranking_100_example.log &#model_bert,fsdp(ddp) CUDA_VISIBLE_DEVICES="0" nohup accelerate launch --config_file ../../../config/default_fsdp.yaml train_reranker.py \#--model_name_or_path "hfl/chinese-roberta-wwm-ext" >./logs/t2ranking_100_example.log & |
Beta Was this translation helpful? Give feedback.
Replies: 2 comments 1 reply
-
后面有问题,请在issue区提问,谢谢。 |
Beta Was this translation helpful? Give feedback.
https://github.com/NLPJCL/RAG-Retrieval/tree/master/rag_retrieval/train/reranker#:~:text=xlmroberta_default_config%2C%20%E5%A6%82%E6%9E%9C%E8%A6%81%E5%9C%A8bge%2Dreranker%2Dbase%E5%92%8Cbce%2Dreranker%2Dbase_v1%E7%9A%84%E5%9F%BA%E7%A1%80%E4%B8%8A%E8%BF%9B%E8%A1%8C%E5%BE%AE%E8%B0%83%EF%BC%8C%E9%87%87%E7%94%A8%E8%AF%A5%E9%85%8D%E7%BD%AE%E6%96%87%E4%BB%B6%EF%BC%8C%E5%9B%A0%E4%B8%BA%E4%B8%A4%E8%80%85%E5%9C%A8%E5%A4%9A%E8%AF%AD%E8%A8%80%E7%9A%84xlmroberta%E7%9A%84%E5%9F%BA%E7%A1%80%E4%B8%8A%E8%AE%AD%E7%BB%83%E8%80%8C%E6%9D%A5%E3%80%82