Please provide the correct description of the deployment information in the README.
I have to say that the professionalism of your team really needs improvement! The model deployment address described in your README document does not match the actual address read in the code! Although users with development experience can easily solve the problem based on the error reports, for a development team, such consistency of information is a fundamental professional quality. Wan2.1 is a great model, but the professional attitude of Pai Studio is really a major drawback.
Sorry, the person responsible for uploading the model, training the model, and writing the README is all the same person 😂. Sometimes I can't manage everything simultaneously. Could you please let me know where the mistakes are, and I'll make the corrections?
In your README document, it is described that when deploying with Comfy, the model should be placed in the "models/diffusion_transform" directory. However, in reality, in your official workflow, the "nodes.py" file under the "custom_nodes\VideoX-Fun\comfyui\wan2_1_fun" directory has hardcoded the path to look for the model under "CogVideoX_Fun", "Fun_Models", and "VideoX_Fun". Moreover, this deployment method will result in the deployment of files such as CLIP and VAE, which are about 10GB in size, twice, even though they can clearly be shared by the InP and control models.
Sorry for the confusion in our explanation. We provide our own UI, which requires reading files from the models/Diffusion_Transformer directory. ComfyUI is supported separately, so it is placed elsewhere. We will make some modifications to explain things more clearly.