illustration-HD

模型介绍

本模型依托魔搭社区（ModelScope）AIGC专区模型训练环境与算力完成训练。

• 模型类型：LoRA
• 基础模型：HiDream-ai/HiDream-O1-Image
• 训练代码：DiffSynth-Studio
• 训练数据量：300
• 总训练步数：10000
• 开源协议：Apache-2.0

推理代码

安装 DiffSynth-Studio：

pip install diffsynth

开始推理：

from diffsynth.pipelines.hidream_o1_image import HiDreamO1ImagePipeline, ModelConfig
import torch

pipe = HiDreamO1ImagePipeline.from_pretrained(
    torch_dtype=torch.bfloat16,
    device="cuda",
    model_configs=[
        ModelConfig(model_id="HiDream-ai/HiDream-O1-Image", origin_file_pattern="*.safetensors"),
    ],
    processor_config=ModelConfig(model_id="HiDream-ai/HiDream-O1-Image"),
)
pipe.load_lora(pipe.dit, ModelConfig(model_id="Devilworld/illustration-HD", origin_file_pattern="illustration-HD_c1-st10000.safetensors"))
prompt = "a cat"
image = pipe(prompt=prompt, num_inference_steps=30, cfg_scale=4)
image.save("image.jpg")