Yuxuan Zhang

We are working on the diffusers library, but it should be noted that currently we can only complete the lora fine-tuning. SFT requires a large amount of video memory, and...

We indeed have not carried out similar operations, but it should be noted that CogVideoX-5b-I2V uses ddpm instead of ddim.

Specifically, you can see the configuration we used to infer CogVideoX5B in the CLI demo and the readme. About DDIM reverse reconstruction video. We indeed have not conducted any related...

这段历史已经传到模型了吗，看一下完整拼接完的提示词

We are still experimenting, and it will definitely OOM at the moment, unable to fine-tune on an 8-card H100 device, we are working on it

maybe save video cause error. Did you cause this every time?

get, can you help make a PR, this is a issue and need to fix

I’m currently conducting research and scheduling, and I will look into potential deployment options next week. At present, this model does not support int4, which could be a significant issue.

想知道您使用了多少数据进行微调，推荐使用100条相似的视频，以及， 您使用了默认配置吗，能提供一下loss的下降情况吗

这个 + 1 不是你准备数据哟用的，你准备数据就是8N。比如48帧 6秒