Update app.py

app.py

@@ -15,15 +15,14 @@ logger = logging.getLogger(__name__)
 # Define output JSON file path
 OUTPUT_JSON_PATH = "captions.json"
 
-# Load the model and processor with memory optimizations
+# Load the model and processor
 MODEL_PATH = "fancyfeast/llama-joycaption-beta-one-hf-llava"
 try:
     processor = AutoProcessor.from_pretrained(MODEL_PATH)
     model = LlavaForConditionalGeneration.from_pretrained(
         MODEL_PATH,
         torch_dtype=torch.float32,  # CPU-compatible dtype
-        low_cpu_mem_usage=True,    # Minimize memory usage
-        load_in_4bit=True          # Enable 4-bit quantization
+        low_cpu_mem_usage=True     # Minimize memory usage
     ).to("cpu")
     model.eval()
     logger.info("Model and processor loaded successfully.")
@@ -71,7 +70,7 @@ def generate_caption(input_image: Image.Image, caption_type: str = "descriptive"
     
     try:
         # Resize image to reduce memory usage
-        input_image = input_image.resize((256, 256))  # Smaller resolution
+        input_image = input_image.resize((256, 256))  # Small resolution
         
         # Prepare the prompt
         prompt = f"Write a {caption_length} {caption_type} caption for this image."