Upload 3 files

Dockerfile

@@ -0,0 +1,43 @@
+# ─── Base Image ────────────────────────────────────────────────────
+FROM python:3.10-slim
+
+# ─── متغيرات البيئة ────────────────────────────────────────────────
+ENV PYTHONDONTWRITEBYTECODE=1 \
+    PYTHONUNBUFFERED=1 \
+    HF_HOME=/app/.cache/huggingface \
+    TRANSFORMERS_CACHE=/app/.cache/huggingface \
+    PORT=7860
+
+# ─── مجلد العمل ───────────────────────────────────────────────────
+WORKDIR /app
+
+# ─── تثبيت dependencies النظام ────────────────────────────────────
+RUN apt-get update && apt-get install -y \
+    git \
+    wget \
+    libglib2.0-0 \
+    libsm6 \
+    libxext6 \
+    libxrender-dev \
+    libgomp1 \
+    && rm -rf /var/lib/apt/lists/*
+
+# ─── نسخ requirements أولاً (cache optimization) ──────────────────
+COPY requirements.txt .
+
+# ─── تثبيت Python packages ────────────────────────────────────────
+RUN pip install --no-cache-dir --upgrade pip && \
+    pip install --no-cache-dir -r requirements.txt
+
+# ─── نسخ الكود ────────────────────────────────────────────────────
+COPY app.py .
+
+# ─── إنشاء مجلد الكاش ─────────────────────────────────────────────
+RUN mkdir -p /app/.cache/huggingface && \
+    chmod -R 777 /app/.cache
+
+# ─── المنفذ ────────────────────────────────────────────────────────
+EXPOSE 7860
+
+# ─── تشغيل التطبيق ────────────────────────────────────────────────
+CMD ["python", "app.py"]

app.py

@@ -0,0 +1,129 @@
+# ─── flash_attn Mock ─────────────────────────────────────────────
+import sys
+import types
+import importlib.util
+
+flash_mock = types.ModuleType("flash_attn")
+flash_mock.__version__ = "2.0.0"
+flash_mock.__spec__ = importlib.util.spec_from_loader("flash_attn", loader=None)
+sys.modules["flash_attn"] = flash_mock
+sys.modules["flash_attn.flash_attn_interface"] = types.ModuleType("flash_attn.flash_attn_interface")
+sys.modules["flash_attn.bert_padding"] = types.ModuleType("flash_attn.bert_padding")
+# ─────────────────────────────────────────────────────────────────
+
+import io
+import time
+import torch
+from PIL import Image
+from transformers import AutoProcessor, AutoModelForCausalLM
+from fastapi import FastAPI, HTTPException, UploadFile, File
+from contextlib import asynccontextmanager
+
+MODEL_ID = "microsoft/Florence-2-large-ft"
+
+# ─── السؤال الأصلي + تأكيد على اليد ─────────────────────────────
+VQA_QUESTION = (
+    "Is there a woman or any part of a woman's body in this image? Answer yes or no only."
+)
+
+MODEL_DATA = {}
+
+@asynccontextmanager
+async def lifespan(app: FastAPI):
+    print(f"📥 Loading {MODEL_ID}...")
+    start = time.time()
+    MODEL_DATA["processor"] = AutoProcessor.from_pretrained(
+        MODEL_ID, trust_remote_code=True
+    )
+    MODEL_DATA["model"] = AutoModelForCausalLM.from_pretrained(
+        MODEL_ID,
+        torch_dtype=torch.float32,
+        trust_remote_code=True,
+        attn_implementation="eager"
+    ).eval()
+    print(f"✅ Model ready in {time.time()-start:.1f}s")
+    yield
+    MODEL_DATA.clear()
+
+app = FastAPI(
+    title="Female Detection API - VQA",
+    description="Florence-2-large-ft | VQA",
+    version="4.3.0",
+    lifespan=lifespan
+)
+
+@app.get("/health")
+def health():
+    return {"status": "ok", "model_loaded": "model" in MODEL_DATA}
+
+def decide(answer: str) -> tuple[str, str]:
+    """
+    - "no"        → allow  ✅
+    - "yes"       → block  🔴
+    - أي شيء آخر → block  🔴 للأمان
+    """
+    a = answer.strip().lower()
+    if a == "no" or a.startswith("no"):
+        return "allow", "model_answered_no"
+    elif "yes" in a:
+        return "block", "model_answered_yes"
+    else:
+        return "block", "unexpected_answer_blocked_for_safety"
+
+@app.post("/analyze")
+async def analyze_image(file: UploadFile = File(...)):
+
+    if not file.content_type.startswith("image/"):
+        raise HTTPException(status_code=400, detail="الملف ليس صورة")
+
+    try:
+        image = Image.open(io.BytesIO(await file.read())).convert("RGB")
+    except Exception as e:
+        raise HTTPException(status_code=400, detail=f"خطأ في قراءة الصورة: {str(e)}")
+
+    try:
+        processor = MODEL_DATA["processor"]
+        model     = MODEL_DATA["model"]
+
+        task   = "<VQA>"
+        prompt = f"{task}{VQA_QUESTION}"
+
+        inputs = processor(text=prompt, images=image, return_tensors="pt")
+
+        start_time = time.time()
+        with torch.no_grad():
+            generated_ids = model.generate(
+                input_ids=inputs["input_ids"],
+                pixel_values=inputs["pixel_values"],
+                max_new_tokens=10,
+                num_beams=3,
+                do_sample=False
+            )
+
+        generated_text = processor.batch_decode(generated_ids, skip_special_tokens=False)[0]
+        parsed  = processor.post_process_generation(
+            generated_text,
+            task=task,
+            image_size=(image.width, image.height)
+        )
+        elapsed = round(time.time() - start_time, 2)
+
+        answer           = parsed.get(task, "").strip()
+        decision, reason = decide(answer)
+
+        return {
+            "decision":       decision,
+            "reason":         reason,
+            "vqa_answer":     answer,
+            "question":       VQA_QUESTION,
+            "execution_time": elapsed,
+            "status":         "success"
+        }
+
+    except Exception as e:
+        raise HTTPException(status_code=500, detail=str(e))
+
+
+if __name__ == "__main__":
+    import uvicorn
+    uvicorn.run(app, host="0.0.0.0", port=7860)

requirements.txt

@@ -0,0 +1,11 @@
+transformers==4.40.0
+timm
+einops
+pillow
+requests
+fastapi
+uvicorn
+pydantic
+torch
+torchvision
+python-multipart