Yuxuan Zhang

不支持zero3，相关的修改我们还没改，支持多机多卡，但是每张卡需要70G以上显存

直接用中文模型微调，支持中文的

更新了最新的hf 代码和微调代码了吗，发个配置？我这边微调是正常的

cuda12.3 记得拉最新的hf中的modeling_cogvlm

Can you check the latest main branch? I believe test_inference has not been uploaded, and using the new inference directly in the current main branch should work.

Did the out-of-memory error occur at 10 seconds, but not at 5 seconds? I encountered this problem while running it today as well. Next week, our submission of the diffusers...

When fine-tuning, this is normal because fakecp only optimizes the VAE decoder part without touching the VAE encoder part. We haven't paid much attention to this issue before, we will...

Currently, if you can preprocess the data in advance instead of following the current code implementation, it will significantly improve GPU performance, as a lot of time is wasted in...

However, this plan is not suitable for those who want to start training immediately, but rather for enterprise teams that require systematic fine-tuning. Therefore, we will not update this part...

我们的模型不支持负面提示词，所以我才写了全0的，加了也不行