Skip to content

Conversation

@dyhuachi
Copy link
Contributor

@dyhuachi dyhuachi commented Dec 5, 2025

✨ 功能说明

新增 scripts/merge_lora_model.py 脚本,支持将训练好的 LoRA 权重合并到基础 MiniMind 模型中,生成完整可部署的模型。
新增 scripts/call_merged_model.py 脚本,支持测试推理合并后的模型

🔧 使用方式

python scripts/merge_lora.py --lora_model_path ... --base_model_path ... --output_path ...
<img width="1510" height="789" alt="image" src="https://github.com/user-attachments/assets/a7a6a3c9-b175-439d-8dbb-6e9580d533b8" />

@dyhuachi
Copy link
Contributor Author

dyhuachi commented Dec 5, 2025

因为embedding和lm_head共享权重,需要单独保存才能使用safetensor,修改之后对应修改
# 保存模型(使用 safe_serialization=True 保存为 safetensors)
model.save_pretrained(output_path, safe_serialization=False)为true,如果通过评审,可以讨论一下保存模型权重的时候是否增加参数单独保存lm_head部分的权重(默认值可以为false)做兼容性扩展

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant