Question about training and inference

[baolef]

Hi, I'm trying to train DaS with my own data. scripts/train_DaS.sh is used to train the model, where I set MODEL_PATH="checkpoints/Diffusion-As-Shader" and MODEL_PATH="checkpoints/CogVideoX-5b-I2V" for two starting point. After training, I noticed that only the transformer is saved. Therefore, I replace the original transformer with the saved one for inference using demo.py. Is my procedure correct?

[skygoo2000]

Sorry for responding so late.
Yes, it's correct