Force CUDA use

Dockerfile

@@ -1,63 +1,41 @@
-FROM python:3.10-slim
+FROM nvidia/cuda:12.4.1-runtime-ubuntu22.04
 
-# Add CUDA repository and install CUDA toolkit
-RUN apt-get update && apt-get install -y --no-install-recommends \
-    wget \
-    gnupg2 \
-    && wget -qO - https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/3bf863cc.pub | apt-key add - \
-    && echo "deb https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64 /" > /etc/apt/sources.list.d/cuda.list \
-    && apt-get update \
-    && apt-get install -y --no-install-recommends \
-    cuda-toolkit-12-4 \
-    libcudnn8 \
-    libcublas-12-4 \
-    && rm -rf /var/lib/apt/lists/*
-
-# Install build tools and system dependencies
+# Install Python 3.10 and essential dependencies
 RUN apt-get update && apt-get install -y --no-install-recommends \
+    python3.10 \
+    python3.10-dev \
+    python3-pip \
     build-essential \
-    gcc \
-    g++ \
     git \
+    libgl1-mesa-glx \
     libglib2.0-0 \
     libsm6 \
     libxext6 \
     libxrender-dev \
-    elfutils \
     && rm -rf /var/lib/apt/lists/*
 
-# Verify Python version
-RUN python --version && pip --version
+# Set Python 3.10 as default
+RUN update-alternatives --install /usr/bin/python python /usr/bin/python3.10 1 && \
+    update-alternatives --install /usr/bin/pip pip /usr/bin/pip3 1
 
-WORKDIR /code
+WORKDIR /app
 
-# Copy requirements first for better layer caching
+# Copy and install requirements
 COPY requirements.txt .
+RUN pip install --no-cache-dir --upgrade pip && \
+    pip install --no-cache-dir -r requirements.txt
 
-# Install Python packages with pip upgrade
-RUN python3.10 -m pip install --no-cache-dir --upgrade pip setuptools wheel && \
-    python3.10 -m pip install --no-cache-dir -r requirements.txt
-
-# Fix executable stack issue in onnxruntime binary
-RUN find /usr/local/lib/python3.10/site-packages/onnxruntime -name "*.so" -type f 2>/dev/null | while read f; do \
-    execstack -c "$f" 2>/dev/null || true; \
-done || true
-
-# Fix CUDA library symlinks for onnxruntime compatibility
-RUN find /usr/local/cuda/lib64 -name "libcublas*" -o -name "libcudnn*" 2>/dev/null && \
-    ln -sf /usr/local/cuda/lib64/libcublasLt.so.12.* /usr/local/cuda/lib64/libcublasLt.so.12 2>/dev/null || true && \
-    ln -sf /usr/local/cuda/lib64/libcublas.so.12.* /usr/local/cuda/lib64/libcublas.so.12 2>/dev/null || true && \
-    ln -sf /usr/local/cuda/lib64/libcudart.so.12.* /usr/local/cuda/lib64/libcudart.so.12 2>/dev/null || true
-
+# Copy application code
 COPY . .
 
-# Set comprehensive environment variables for CUDA
+# Create insightface cache directory
+RUN mkdir -p /tmp/.insightface
+
+# Set environment variables
 ENV CUDA_HOME=/usr/local/cuda
+ENV LD_LIBRARY_PATH=${CUDA_HOME}/lib64:${LD_LIBRARY_PATH}
 ENV PATH=${CUDA_HOME}/bin:${PATH}
-ENV LD_LIBRARY_PATH=${CUDA_HOME}/lib64:${CUDA_HOME}/extras/CUPTI/lib64:/usr/local/cuda/lib64:/usr/lib/x86_64-linux-gnu:${LD_LIBRARY_PATH}
 ENV INSIGHTFACE_HOME=/tmp/.insightface
-ENV MPLCONFIGDIR=/tmp/matplotlib
-ENV ORT_CUDA_UNAVAILABLE=0
 
 EXPOSE 7860
 

handler.py

@@ -19,8 +19,16 @@ from concurrent.futures import ThreadPoolExecutor, as_completed
 
 class EndpointHandler:
     def __init__(self, model_dir=None):
-        self.app = FaceAnalysis(root="/tmp/.insightface")
-        self.app.prepare(ctx_id=0)  # Set to 0 for GPU, or -1 for CPU
+        # Initialize FaceAnalysis with GPU support
+        self.app = FaceAnalysis(root="/tmp/.insightface", providers=['CUDAExecutionProvider', 'CPUExecutionProvider'])
+        self.app.prepare(ctx_id=0)  # 0 = GPU, -1 = CPU
+        
+        print("=" * 80)
+        print("InsightFace Providers:")
+        for model in self.app.models:
+            if hasattr(model, 'sess'):
+                print(f"  {model.__class__.__name__}: {model.sess.get_providers()}")
+        print("=" * 80)
         
         # Get configuration
         config = get_config()