sd/model_loader.py

from sd.clip import CLIP
from sd.encoder import VAE_Encoder
from sd.decoder import VAE_Decoder
from sd.diffusion import Diffusion
from sd import model_converter

def from_pretrained(weights_url, device):
    state_dict = model_converter.from_pretrained(weights_url, device)

    encoder = VAE_Encoder().to(device)
    encoder.load_state_dict(state_dict['encoder'], strict=True)
    
    decoder = VAE_Decoder().to(device)
    decoder.load_state_dict(state_dict['decoder'], strict=True)

    diffusion = Diffusion().to(device)
    diffusion.load_state_dict(state_dict['diffusion'], strict=True)

    clip = CLIP().to(device)
    clip.load_state_dict(state_dict['clip'], strict=True)

    return {
        'clip': clip,
        'encoder': encoder,
        'decoder': decoder,
        'diffusion': diffusion,
    }