Training/Fine-tuning Instructions

[JhEdLp]

Hi,

Thank you for developing this interesting project.

I'm writing to request information on how to train or fine-tune the model. Could you please provide the steps involved, including details on dataset creation or preparation?

Thanks in advance for your assistance.

[zhuangshaobin]

Training Showmaker is a relatively straightforward process. Currently, you can use frameworks like Accelerate and DeepSpeed to train the model either on a single GPU or in a distributed manner. For SFT, you only need to prepare 1,000+ high-quality video data samples.