通义千问本地配置并实现微调

最小Qwen模型大小942mb

from modelscope import snapshot_download
model_dir = snapshot_download(“qwen/Qwen2.5-0.5B”, cache_dir=“./models2.5”)

Qwen2.5-0.5B：942MB

from modelscope import snapshot_download
model_dir = snapshot_download(“qwen1/Qwen1.5-1.8B”, cache_dir=“./models”)

Qwen1.5-1.8B：2.88GB

一、环境准备

1. 硬件要求