模型： qwen2.5:14b-instruct qwen2.5:7b-instruct