it works

Dockerfile

@@ -0,0 +1,28 @@
+FROM pytorch/pytorch:2.2.1-cuda12.1-cudnn8-runtime
+
+# Install git, git-lfs and basic dependencies
+RUN apt-get update && apt-get install -y git git-lfs curl ffmpeg libsm6 libxext6 libgl1-mesa-glx xvfb
+
+# Install Python dependencies including OpenCV
+RUN pip install --no-cache-dir runpod requests opencv-python-headless
+
+# Set up environment variables for HuggingFace and torch cache
+ENV HUGGINGFACE_HUB_CACHE=/workspace/cache
+ENV TORCH_HOME=/workspace/models
+ENV TRANSFORMERS_CACHE=/workspace/cache
+ENV XDG_CACHE_HOME=/workspace/cache
+ENV PYTHONUNBUFFERED=1
+
+# Copy handler and startup script
+COPY rp_handler.py /
+COPY start.sh /
+RUN chmod +x /start.sh
+
+# Create directory stubs for workspace to avoid permission issues
+RUN mkdir -p /workspace/app /workspace/cache /workspace/models
+
+# Set the workspace as working directory
+WORKDIR /
+
+# Start with our script
+CMD ["/start.sh"] 

rp_handler.py

@@ -0,0 +1,94 @@
+import runpod
+import io
+import base64
+import time
+import requests
+import sys
+import os
+import traceback
+from PIL import Image
+import torch
+
+def download_image(url):
+    """Download an image from a URL and return a PIL Image object."""
+    try:
+        response = requests.get(url, stream=True)
+        response.raise_for_status()
+        return Image.open(io.BytesIO(response.content))
+    except Exception as e:
+        print(f"Error downloading image from {url}: {str(e)}")
+        raise
+
+def encode_image_to_base64(image):
+    """Encode a PIL Image to base64 string."""
+    buffered = io.BytesIO()
+    image.save(buffered, format="PNG")
+    img_str = base64.b64encode(buffered.getvalue()).decode()
+    return img_str
+
+def handler(event):
+    try:
+        print("Handler started, importing inference module...")
+        from inference_utils import inference
+        
+        start_time = time.time()
+        print("Processing request...")
+        
+        # Extract input data
+        input_data = event["input"]["data"]
+        
+        if len(input_data) < 3:
+            return {
+                "status": "error",
+                "message": "Missing required parameters. Expected [id_image_url, makeup_image_url, guidance_scale]"
+            }
+        
+        id_image_url = input_data[0]
+        makeup_image_url = input_data[1]
+        guidance_scale = float(input_data[2]) if len(input_data) > 2 else 1.6
+        
+        print(f"Downloading images from URLs...")
+        id_image = download_image(id_image_url)
+        makeup_image = download_image(makeup_image_url)
+        
+        print(f"Running inference with guidance scale {guidance_scale}...")
+        result_image = inference(id_image, makeup_image, guidance_scale)
+        
+        # Calculate processing time
+        processing_time = time.time() - start_time
+        print(f"Processing completed in {processing_time:.2f} seconds")
+        
+        # Return base64 encoded image
+        return {
+            "status": "completed",
+            "image": encode_image_to_base64(result_image),
+            "processingTime": processing_time
+        }
+    
+    except Exception as e:
+        # Print full exception for debugging
+        print(f"Error in handler: {str(e)}")
+        print(traceback.format_exc())
+        
+        # Clean up GPU memory
+        if torch.cuda.is_available():
+            torch.cuda.empty_cache()
+            
+        return {
+            "status": "error",
+            "message": str(e)
+        }
+
+if __name__ == "__main__":
+    print(f"Starting RunPod Serverless handler from {os.getcwd()}")
+    print(f"Python version: {sys.version}")
+    print(f"CUDA available: {torch.cuda.is_available()}")
+    if torch.cuda.is_available():
+        print(f"CUDA device: {torch.cuda.get_device_name(0)}")
+        
+    # Check if all environment variables are set
+    print(f"HF cache: {os.environ.get('HUGGINGFACE_HUB_CACHE')}")
+    print(f"Torch home: {os.environ.get('TORCH_HOME')}")
+    
+    # Start the handler
+    runpod.serverless.start({"handler": handler}) 

start.sh

@@ -0,0 +1,66 @@
+#!/bin/bash
+set -e
+
+echo "Starting RunPod Stable-Makeup Serverless"
+
+# Check if workspace volume is properly mounted
+if [ ! -d "/workspace" ]; then
+    echo "ERROR: /workspace volume is not mounted. Please create a RunPod with a Network Volume mounted at /workspace"
+    exit 1
+fi
+
+# Check available disk space and display
+df -h /workspace
+df -h /
+
+# Make sure our directories exist
+mkdir -p /workspace/app /workspace/cache /workspace/models /workspace/models/stablemakeup
+
+# Clone the repository with LFS since we have sufficient network storage (40GB)
+if [ ! -d "/workspace/app/stable-makeup" ]; then
+    echo "Cloning repository with LFS from Hugging Face..."
+    
+    # Setup git-lfs
+    git lfs install
+    
+    # Clone the repository with LFS
+    git clone https://huggingface.co/spaces/edgarhnd/Stable-Makeup-unofficial /workspace/app/stable-makeup
+    
+    # Copy our handler file
+    cp /rp_handler.py /workspace/app/stable-makeup/
+    
+    # Modify the inference_utils.py file to use the workspace volume for checkpoints
+    sed -i 's|"./checkpoints/stablemakeup"|"/workspace/models/stablemakeup"|g' /workspace/app/stable-makeup/inference_utils.py
+    
+    # Install required dependencies
+    cd /workspace/app/stable-makeup
+    echo "Installing dependencies..."
+    pip install --no-cache-dir -r requirements.txt
+    pip install --no-cache-dir opencv-python-headless
+    
+    # Create checkpoints directory and copy model files if they exist in the repo
+    mkdir -p /workspace/models/stablemakeup
+    if [ -d "/workspace/app/stable-makeup/checkpoints/stablemakeup" ]; then
+        echo "Copying model files from cloned repository to workspace models directory..."
+        cp -v /workspace/app/stable-makeup/checkpoints/stablemakeup/* /workspace/models/stablemakeup/
+    fi
+fi
+
+# If model files are still missing, download them directly
+if [ ! -f "/workspace/models/stablemakeup/pytorch_model.bin" ]; then
+    echo "Model files not found in cloned repository. Downloading directly..."
+    
+    echo "Downloading pytorch_model.bin..."
+    curl -L -o /workspace/models/stablemakeup/pytorch_model.bin https://huggingface.co/spaces/edgarhnd/Stable-Makeup-unofficial/resolve/main/checkpoints/stablemakeup/pytorch_model.bin
+    
+    echo "Downloading pytorch_model_1.bin..."
+    curl -L -o /workspace/models/stablemakeup/pytorch_model_1.bin https://huggingface.co/spaces/edgarhnd/Stable-Makeup-unofficial/resolve/main/checkpoints/stablemakeup/pytorch_model_1.bin
+    
+    echo "Downloading pytorch_model_2.bin..."
+    curl -L -o /workspace/models/stablemakeup/pytorch_model_2.bin https://huggingface.co/spaces/edgarhnd/Stable-Makeup-unofficial/resolve/main/checkpoints/stablemakeup/pytorch_model_2.bin
+fi
+
+# Start the handler
+echo "Starting serverless handler..."
+cd /workspace/app/stable-makeup
+exec python -u rp_handler.py 

test_input.json

@@ -0,0 +1,9 @@
+{
+  "input": {
+    "data": [
+      "https://cool-dodo-849.convex.cloud/api/storage/488bd925-6741-4135-89e7-3ae2bac78e91",
+      "https://cool-dodo-849.convex.cloud/api/storage/2aa3db4c-2cd0-4d5e-b84a-ea0a7b6f3615",
+      1.6
+    ]
+  }
+}