我有个类似问题。假定基础模型Model 1.我在基础模型Model 的基础上用数据集A微调了LoRA权重LoRA_A 2.我想继续在LoRA权重A的基础上,用数据集B微调LoRA权重LoRA_B 我不希望将LoRA_A存成一个基础模型,我希望直接从LoRA_A的权重微调LoRA_B,获得的B_adapter直接能用在原始模型Model的权重上,我该如何操作? #6220
Unanswered
DataNinja42298
asked this question in
Q&A
Replies: 2 comments
-
训完A_lora之后,与base模型合并,再基于该模型微调B_lora。最终B_lora和base模型合并。这样是不是满足你的需要了? |
Beta Was this translation helpful? Give feedback.
0 replies
-
直接将A_adaptor和B_adaptor的子矩阵权重分别相乘,得到同维度矩阵,再按照element-wise相加,就得到C_adaptor,这就可以用在跟原始权重合并了。你想基于C_adaptor获得低秩矩阵,只能做SGD分解了。 |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Reminder
System Info
我有个类似问题。假定基础模型Model
1.我在基础模型Model 的基础上用数据集A微调了LoRA权重LoRA_A
2.我想继续在LoRA权重A的基础上,用数据集B微调LoRA权重LoRA_B
我不希望将LoRA_A存成一个基础模型,我希望直接从LoRA_A的权重微调LoRA_B,获得的B_adapter直接能用在原始模型Model的权重上,我该如何操作?
Reproduction
我有个类似问题。假定基础模型Model
1.我在基础模型Model 的基础上用数据集A微调了LoRA权重LoRA_A
2.我想继续在LoRA权重A的基础上,用数据集B微调LoRA权重LoRA_B
我不希望将LoRA_A存成一个基础模型,我希望直接从LoRA_A的权重微调LoRA_B,获得的B_adapter直接能用在原始模型Model的权重上,我该如何操作?
Expected behavior
No response
Others
No response
Beta Was this translation helpful? Give feedback.
All reactions