from diffsynth.pipelines.flux3_image import Flux3ImagePipeline, ModelConfig from diffsynth.core import load_state_dict import torch pipe = Flux3ImagePipeline.from_pretrained( torch_dtype=torch.bfloat16, device="cuda", model_configs=[ ModelConfig(model_id="black-forest-labs/flux-3-image-klein-base", origin_file_pattern="model.safetensors"), ModelConfig(model_id="black-forest-labs/flux-3-image-klein", origin_file_pattern="text_encoder/*.safetensors"), ModelConfig(model_id="black-forest-labs/flux-3-image-klein", origin_file_pattern="vae/diffusion_pytorch_model.safetensors"), ], processor_config=ModelConfig(model_id="black-forest-labs/flux-3-image-klein", origin_file_pattern="text_encoder/"), ) state_dict = load_state_dict("./models/train/flux-3-image-klein-base_full/epoch-1.safetensors", torch_dtype=torch.bfloat16) pipe.dit.load_state_dict(state_dict) prompt = "a dog" image = pipe(prompt=prompt, seed=0, num_inference_steps=50, cfg_scale=3.5) image.save("image.jpg")