Upload 3 files

DockerFile

@@ -0,0 +1,11 @@
+FROM python:3.10-slim
+
+WORKDIR /app
+
+COPY requirements.txt requirements.txt
+RUN pip install --no-cache-dir -r requirements.txt
+
+COPY app.py app.py
+
+EXPOSE 7860
+CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860"]

app.py

@@ -0,0 +1,47 @@
+from fastapi import FastAPI, UploadFile, Form
+from PIL import Image
+from gui_actor.modeling import Qwen2VLForConditionalGenerationWithPointer
+from transformers import Qwen2VLProcessor
+from gui_actor.inference import inference
+import torch
+import io
+
+app = FastAPI()
+
+# Load model + processor at startup
+MODEL_NAME = "microsoft/GUI-Actor-2B-Qwen2-VL"
+processor = Qwen2VLProcessor.from_pretrained(MODEL_NAME)
+tokenizer = processor.tokenizer
+model = Qwen2VLForConditionalGenerationWithPointer.from_pretrained(
+    MODEL_NAME,
+    torch_dtype=torch.float32,
+    device_map="auto"
+).eval()
+
+@app.get("/")
+def home():
+    return {"message": "GUI-Actor Space is running"}
+
+@app.post("/predict/")
+async def predict(
+    instruction: str = Form(...),
+    image: UploadFile = Form(...)
+):
+    # Read and process image
+    img_bytes = await image.read()
+    img = Image.open(io.BytesIO(img_bytes)).convert("RGB")
+    
+    # Auto resize if needed
+    max_width, max_height = 480, 270
+    if img.width > max_width or img.height > max_height:
+        img.thumbnail((max_width, max_height))
+    
+    # Run inference
+    click_point = inference(
+        instruction=instruction,
+        image=img,
+        model=model,
+        processor=processor,
+        tokenizer=tokenizer
+    )
+    return {"click_point": click_point}

requirements.txt

@@ -0,0 +1,7 @@
+fastapi
+uvicorn
+pillow
+torch
+transformers
+qwen-vl-utils
+git+https://github.com/microsoft/GUI-Actor.git