app.py

import streamlit as st
import io
import requests
from PIL import Image
from io import BytesIO

# ----------------------------
# 配置 Hugging Face Inference API
# ----------------------------
import os




# API_URL = "https://api-inference.huggingface.co/models/GGPENG/StyleDiffusion"  # 替换为你上传的模型仓库
# API_TOKEN = os.getenv("HF_TOKEN")

# headers = {"Authorization": f"Bearer {API_TOKEN}"}

# # ----------------------------
# # Streamlit 页面设置
# # ----------------------------
# st.set_page_config(page_title="Fine-tuning style diffusion (API)", layout="wide")

# st.title("Fine-tuning style diffusion 推理 Demo (API)")
# st.write("只是训练了一个提示词 'A <new1> reference.'")
# st.write("示例：A <new1> reference. New Year image with a rabbit as the main element, in a 2D or anime style, and a festive background")

# # ----------------------------
# # Prompt 输入
# # ----------------------------
# prompt = st.text_input(
#     "Prompt",
#     "A <new1> reference. New Year image with a rabbit as the main element, in a 2D or anime style, and a festive background"
# )

# # ----------------------------
# # 参数调节
# # ----------------------------
# steps = st.slider("Steps", 10, 320, 100)
# guidance = st.slider("Guidance", 1.0, 18.0, 6.0)

# # ----------------------------
# # 生成函数（调用 API）
# # ----------------------------
# def generate(prompt):
#     payload = {
#         "inputs": prompt,
#         "parameters": {
#             "num_inference_steps": steps,
#             "guidance_scale": guidance,
#             # "height": 512,
#             # "width": 512
#         }
#     }

#     response = requests.post(API_URL, headers=headers, json=payload)
    
#     if response.status_code != 200:
#         st.error(f"API请求失败：{response.status_code}, {response.text}")
#         return None
    
#     # 将返回的字节流或 Base64 数据转换为 PIL Image
#     try:
#         image = Image.open(BytesIO(response.content))
#     except:
#         st.error("生成图像失败，请检查模型是否支持图像输出。")
#         return None
#     return image

# # ----------------------------
# # 生成按钮
# # ----------------------------
# if st.button("Generate"):
#     with st.spinner("Generating via Hugging Face API..."):
#         image = generate(prompt)
    
#     if image:
#         st.image(image, caption="Result", width=512)
#         buf = io.BytesIO()
#         image.save(buf, format="PNG")
#         st.download_button(
#             "Download",
#             buf.getvalue(),
#             "result.png"
#         )



from diffusers import StableDiffusionPipeline
import torch
from PIL import Image
import io

# ----------------------------
# 加载基础模型
# ----------------------------
base_model = "runwayml/stable-diffusion-v1-5"
pipe = StableDiffusionPipeline.from_pretrained(base_model, torch_dtype=torch.float16)
pipe = pipe.to("cuda")  # 有GPU加速

# ----------------------------
# 加载自定义微调权重
# ----------------------------
ckpt_path = "./pytorch_custom_diffusion_weights.bin"

# 假设你用的是 Diffusers 支持的 UNet 权重增量加载
pipe.unet.load_attn_procs(ckpt_path)


import streamlit as st

st.set_page_config(page_title="Custom Style Diffusion Demo", layout="wide")
st.title("Custom Style Diffusion 本地推理 Demo")

prompt = st.text_input("Prompt", "A <new1> reference. New Year image with a rabbit in 2D anime style")
steps = st.slider("Steps", 10, 320, 50)
guidance = st.slider("Guidance Scale", 1.0, 20.0, 7.5)

if st.button("Generate"):
    with st.spinner("Generating image..."):
        result = pipe(prompt, num_inference_steps=steps, guidance_scale=guidance)
        image = result.images[0]
        st.image(image, caption="Result", use_column_width=True)

        buf = io.BytesIO()
        image.save(buf, format="PNG")
        st.download_button("Download Image", buf.getvalue(), "result.png")