update code

README.md

@@ -1,42 +1,20 @@
-# AI 形象照生成服务 - 基于 FLUX.1-Fill-dev 模型
+# FLUX.1-Fill-dev Custom Inference Endpoint
 
-本项目基于 `black-forest-labs/FLUX.1-Fill-dev` 模型，部署为 Hugging Face Inference Endpoint 服务，支持**图生图（image-to-image / inpainting）**应用场景：
+This repository is set up to deploy a custom Hugging Face Inference Endpoint using the StableDiffusionInstructPix2PixPipeline, enabling image-to-image transformation with prompt guidance.
 
-- 替换背景
-- 风格转换
-- 商务形象照生成
+## How to Use
 
-## 🔧 接口调用说明（POST /）
-
-接口输入格式（JSON）：
+Send a POST request with base64-encoded `image` and your `prompt` to get an enhanced version of the image.
 
+## Input example:
 ```json
 {
   "inputs": {
-    "prompt": "商务风格形象照，穿蓝色西装，微笑，高清写实风格",
-    "image": "<base64 PNG 图像>",
-    "mask": "<base64 PNG 掩码图像>"
+    "prompt": "高清艺术风格头像",
+    "image": "<base64-encoded-image>",
+    "steps": 30,
+    "guidance_scale": 7.5,
+    "image_guidance_scale": 1.5
   }
 }
 ```
-
-返回格式：
-
-```json
-{
-  "image": "<base64 PNG 图像>"
-}
-```
-
-## 💡 掩码说明
-- 图像和掩码需为相同尺寸
-- 掩码中**白色部分为需要替换的区域**
-
-## 🚀 示例场景
-
-- 用户上传自拍照片，并通过 mask 指定区域（如衣服、背景）
-- 后端根据 prompt 生成对应风格的合成图
-
----
-
-本部署服务基于 diffusers 实现，更多信息参考：https://huggingface.co/black-forest-labs/FLUX.1-Fill-dev

handler.py

@@ -1,21 +1,17 @@
-from diffusers import StableDiffusionInpaintPipeline
+from diffusers import StableDiffusionInstructPix2PixPipeline
 from PIL import Image
 import torch
 import base64
 from io import BytesIO
 
-# ✅ 使用 huggingface repo id 形式加载（确保路径正确）
-pipe = StableDiffusionInpaintPipeline.from_pretrained(
-    "shangguanyanyan/flux1-fill-dev-custom",  # 请确认仓库ID无误
+pipe = StableDiffusionInstructPix2PixPipeline.from_pretrained(
+    "shangguanyanyan/flux1-fill-dev-custom",
     torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32
 ).to("cuda" if torch.cuda.is_available() else "cpu")
 
 def decode_image(b64_string):
-    try:
-        image_data = base64.b64decode(b64_string)
-        return Image.open(BytesIO(image_data)).convert("RGB")
-    except Exception as e:
-        raise ValueError(f"解码图像失败: {str(e)}")
+    image_data = base64.b64decode(b64_string)
+    return Image.open(BytesIO(image_data)).convert("RGB")
 
 def encode_image(image):
     buffer = BytesIO()
@@ -25,45 +21,23 @@ def encode_image(image):
 def handler(data):
     try:
         inputs = data.get("inputs", {})
-        prompt = inputs.get("prompt", "高清写实风格人物形象照")
-
+        prompt = inputs.get("prompt", "写实风格形象照")
         image_b64 = inputs.get("image")
-        mask_b64 = inputs.get("mask")
-        if not image_b64 or not mask_b64:
-            raise ValueError("缺少必要的 image 或 mask 参数")
 
-        image = decode_image(image_b64)
-        mask = decode_image(mask_b64)
-
-        # 默认参数（支持调整）
-        height = int(inputs.get("height", image.height))
-        width = int(inputs.get("width", image.width))
-        steps = int(inputs.get("num_inference_steps", 30))
-        cfg_scale = float(inputs.get("guidance_scale", 7.5))
+        if not image_b64:
+            return {"error": "缺少 image 参数", "status": "failed"}
 
-        image = image.resize((width, height))
-        mask = mask.resize((width, height))
+        image = decode_image(image_b64)
 
         result = pipe(
             prompt=prompt,
             image=image,
-            mask_image=mask,
-            height=height,
-            width=width,
-            num_inference_steps=steps,
-            guidance_scale=cfg_scale
+            num_inference_steps=int(inputs.get("steps", 30)),
+            guidance_scale=float(inputs.get("guidance_scale", 7.5)),
+            image_guidance_scale=float(inputs.get("image_guidance_scale", 1.5))
         ).images[0]
 
-        return {
-            "status": "success",
-            "image": encode_image(result),
-            "meta": {
-                "prompt": prompt,
-                "size": f"{width}x{height}",
-                "steps": steps,
-                "cfg_scale": cfg_scale
-            }
-        }
+        return {"image": encode_image(result), "status": "success"}
 
     except Exception as e:
-        return {"status": "failed", "error": str(e)}
+        return {"error": str(e), "status": "failed"}

requirements.txt

@@ -1,5 +1,5 @@
-torch>=2.1.0
-transformers>=4.35.0
-diffusers>=0.26.3
-accelerate>=0.25.0
-Pillow>=10.0.0
+diffusers>=0.14.0
+transformers
+torch
+accelerate
+safetensors