[diffusers][model loading]How to load a custom diffusion in sglang diffusion?

[JohnHerry]

Need help about custom model support in sglang. I have a custom diffuser trained with pytorch and get the pt model file. the model structure and model config are not compitable with the huggingface API. Now I want to use sglang[diffuser] as the inference platform. So I had decomposed my model as a tokenEncoder and DiTModel. then I want to check that using sglang to load the model weights.
But it seems that sglang [diffusion] suppose my model is compitable with huggingface, Or, the tokenEncoder and DiTModel are two models. but in my case they are just in One single model, My question is how can I load my model with sglang? a detailed example is better, thanks.