free

Dockerfile

@@ -10,4 +10,4 @@ COPY --chown=user ./requirements.txt requirements.txt
 RUN pip install --no-cache-dir --upgrade -r requirements.txt
 
 COPY --chown=user . /app
-CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860"]
+CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860"]

app.py

@@ -58,39 +58,7 @@ async def startup_event():
         logger.warning("GPU not available. Using CPU, performance may be slow!")
     
     try:
-        # Try to load the model with more information for debugging
-        logger.info("Attempting to load CSM 1B model...")
-        
-        # Import extra tools that might be needed
-        from huggingface_hub import hf_hub_download
-        import json
-        import os
-        
-        # Try to use an alternative loading method if the direct method fails
-        try:
-            # First attempt with default loading
-            generator = load_csm_1b(device=device)
-        except TypeError as e:
-            if "missing 1 required positional argument: 'config'" in str(e):
-                logger.info("Model requires config. Attempting to load with configuration...")
-                
-                # Try to load the configuration first
-                try:
-                    # The model_path can be model_id or path
-                    model_id = "sesame/csm-1b"
-                    # Try to download and load the config
-                    config_file = hf_hub_download(repo_id=model_id, filename="config.json")
-                    with open(config_file, 'r') as f:
-                        config = json.load(f)
-                    
-                    # Now try loading with config
-                    generator = load_csm_1b(device=device, config=config)
-                except Exception as config_error:
-                    logger.error(f"Failed to load configuration: {str(config_error)}")
-                    raise
-            else:
-                raise
-                
+        generator = load_csm_1b(device="cuda")
         logger.info(f"Model loaded successfully on device: {device}")
     except Exception as e:
         logger.error(f"Could not load model: {str(e)}")

docker-compose.yml

@@ -0,0 +1,23 @@
+services:
+  csm:
+    build:
+      context: .
+      args:
+        - BUILDKIT_INLINE_CACHE=1
+    ports:
+      - '7860:7860'
+    volumes:
+      - ./data:/app/data
+      - ~/.huggingface:/root/.huggingface
+    env_file:
+      - .env
+    deploy:
+      resources:
+        reservations:
+          devices:
+            - driver: nvidia
+              count: 1
+              capabilities: [gpu]
+    environment:
+      - HUGGINGFACE_TOKEN=${HUGGINGFACE_TOKEN}
+      - PYTHONDONTWRITEBYTECODE=1