我用lora训练完之后需要选择一个表现较好的checkpoint，只能逐个export导出模型来测试吗？ #4838
Replies: 3 comments 7 replies
-
|
直接用 |
Beta Was this translation helpful? Give feedback.
-
|
請問能給更多資訊嗎?例如關鍵字或者怎麼使用之類的.... 真心想了解 |
Beta Was this translation helpful? Give feedback.
-
|
可以用weibu那个可视化页面，在train和export那行有一个chat，点进去设定检查点，加载模型，就可以初步进行聊天验证了 |
Beta Was this translation helpful? Give feedback.
-
|
直接base model + 各个检查点的 lora adapter 就可以 |
Beta Was this translation helpful? Give feedback.
-
|
大家好，我想问问，这是我的saves/Llama-2-7B-Chat/lora/train_2025-02-18-11-47-28是adapter_name_or_path，而adapter_name_or_path下面这么多个文件和文件夹，export模型时，加载的checkpoint到底是哪一个呢？代码中哪一块有体现呢？ |
Beta Was this translation helpful? Give feedback.
-
|
一个在检查点文件里.safetensors后缀的文件，
…---原始邮件---
发件人: ***@***.***>
发送时间: 2025年2月18日(周二) 下午2:23
收件人: ***@***.***>;
抄送: ***@***.******@***.***>;
主题: Re: [hiyouga/LLaMA-Factory] 我用lora训练完之后需要选择一个表现较好的checkpoint，只能逐个export导出模型来测试吗？ (Discussion #4838)
大家好，我想问问，这是我的saves/Llama-2-7B-Chat/lora/train_2025-02-18-11-47-28是adapter_name_or_path，而adapter_name_or_path下面这么多个文件和文件夹，export模型时，加载的checkpoint到底是哪一个呢？代码中哪一块有体现呢？
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you commented.Message ID: ***@***.***>
|
Beta Was this translation helpful? Give feedback.
-
|
|
Beta Was this translation helpful? Give feedback.
-
|
这是你训练过程中，根据你设置的保存步数设置的，一般加载就是训练完毕的最后一个 |
Beta Was this translation helpful? Give feedback.
-
|
你好 我想请问一下 如果我想使用webui导出某个中间的checkpoint 是不是只需要在目录中将之后的checkpoint删除后 直接导出可以了 谢谢！！！ |
Beta Was this translation helpful? Give feedback.
-
|
请问如果是vl模型可以这样实现吗？我在用qwen2vl，能否不进行merge然后直接加载adapter和base model进行推理？ |
Beta Was this translation helpful? Give feedback.
-
主要是导出模型占用磁盘空间过大，如何运行的时候加载lora的checkpoint呢？我使用的是baichuan2和qwen2。
Beta Was this translation helpful? Give feedback.
All reactions