mirror of
https://github.com/modelscope/DiffSynth-Studio.git
synced 2026-03-21 16:18:13 +00:00
25 lines
824 B
Python
25 lines
824 B
Python
import torch
|
|
from diffsynth import ModelManager, FluxImagePipeline, download_customized_models
|
|
|
|
prompt = "a beautiful Asian girl."
|
|
|
|
# Generate an image using FLUX.1-dev
|
|
model_manager = ModelManager(torch_dtype=torch.bfloat16, device="cuda", model_id_list=["FLUX.1-dev"])
|
|
pipe = FluxImagePipeline.from_model_manager(model_manager)
|
|
|
|
image = pipe(prompt=prompt, seed=0)
|
|
image.save("image.jpg")
|
|
|
|
# Download and load ArtAug LoRA
|
|
lora_path = download_customized_models(
|
|
model_id="DiffSynth-Studio/ArtAug-lora-FLUX.1dev-v1",
|
|
origin_file_path="merged_lora.safetensors",
|
|
local_dir="models/lora",
|
|
downloading_priority=["ModelScope", "HuggingFace"]
|
|
)[0]
|
|
model_manager.load_lora(lora_path, lora_alpha=1.0)
|
|
|
|
# Generate an image using FLUX.1-dev + ArtAug
|
|
image = pipe(prompt=prompt, seed=0)
|
|
image.save("image_artaug.jpg")
|