THUDM / CodeGeeX2

CodeGeeX2: A More Powerful Multilingual Code Generation Model
https://codegeex.cn
Apache License 2.0
7.55k stars 532 forks source link

非量化版本的codegeex2-6B,16GB显存下最大推理的max_new_tokens约为多少?微调后经常推着推着就OOM了 #218

Open CatYing opened 3 months ago

CatYing commented 3 months ago

as topic temperature=0.2, max_new_tokens=4097

PineREN commented 3 months ago

求问 怎么微调的,是用的chatglmv2的代码进行微调的吗?