微调minicpmv2时cpu占用率超高

modelscope / ms-swift

Use PEFT or Full-parameter to finetune 400+ LLMs or 100+ MLLMs. (LLM: Qwen2.5, Llama3.2, GLM4, Internlm2.5, Yi1.5, Mistral, Baichuan2, DeepSeek, Gemma2, ...; MLLM: Qwen2-VL, Qwen2-Audio, Llama3.2-Vision, Llava, InternVL2, MiniCPM-V-2.6, GLM4v, Xcomposer2.5, Yi-VL, DeepSeek-VL, Phi3.5-Vision, ...)

https://swift.readthedocs.io/zh-cn/latest/Instruction/index.html

Apache License 2.0

4.15k stars 368 forks source link

微调minicpmv2时cpu占用率超高 #1008

Closed strawhatboy closed 2 months ago

strawhatboy commented 5 months ago

Describe the bug 使用swift sft命令指定dataset 和val_dataset 进行微调:

CUDA_VISIBLE_DEVICES=0 swift sft --model_type minicpm-v-v2-chat

20 个核心全部占用60%左右

Your hardware and system info CUDA：12.2 系统：Ubuntu22.04 GPU：RTX4090 Torch：2.2.2

Additional context Add any other context about the problem here(在这里补充其他信息)

tastelikefeet commented 2 months ago

用最新代码试试，感觉不复现的说