Update src/app/response.py

src/app/response.py

@@ -1,59 +1,59 @@
-# Necessary imports
-import sys
-import PIL.Image
-import torch
-import gradio as gr
-import spaces
-
-# Local imports
-from src.config import device, model_name, sampling
-from src.app.model import load_model_and_tokenizer
-from src.logger import logging
-from src.exception import CustomExceptionHandling
-
-
-# Model, tokenizer and processor
-model, tokenizer, processor = load_model_and_tokenizer(model_name, device)
-
-
-@spaces.GPU
-def describe_image(text: str, image: PIL.Image.Image, max_new_tokens: int) -> str:
-    """
-    Generates a response based on the given text and image using the model.
-
-    Args:
-        - text (str): The input text to be processed.
-        - image (PIL.Image.Image): The input image to be processed.
-        - max_new_tokens (int): The maximum number of new tokens to generate.
-
-    Returns:
-        str: The generated response text.
-    """
-    try:
-        # Check if image or text is None
-        if not image or not text:
-            gr.Warning("Please provide an image and a question.")
-
-        # Prepare the inputs
-        text = "answer en " + text
-        inputs = processor(text=text, images=image, return_tensors="pt").to(device)
-
-        # Generate the response
-        with torch.inference_mode():
-            generated_ids = model.generate(
-                **inputs, max_new_tokens=max_new_tokens, do_sample=sampling
-            )
-
-        # Decode the generated response
-        result = processor.batch_decode(generated_ids, skip_special_tokens=True)
-
-        # Log the successful generation of the answer
-        logging.info("Answer generated successfully.")
-
-        # Return the generated response
-        return result[0][len(text) :].lstrip("\n")
-
-    # Handle exceptions that may occur during answer generation
-    except Exception as e:
-        # Custom exception handling
-        raise CustomExceptionHandling(e, sys) from e
+# Necessary imports
+import sys
+import PIL.Image
+import torch
+import gradio as gr
+import spaces
+
+# Local imports
+from src.config import device, model_name, sampling
+from src.app.model import load_model_and_processor
+from src.logger import logging
+from src.exception import CustomExceptionHandling
+
+
+# Model and processor
+model, processor = load_model_and_processor(model_name, device)
+
+
+@spaces.GPU
+def describe_image(text: str, image: PIL.Image.Image, max_new_tokens: int) -> str:
+    """
+    Generates a response based on the given text and image using the model.
+
+    Args:
+        - text (str): The input text to be processed.
+        - image (PIL.Image.Image): The input image to be processed.
+        - max_new_tokens (int): The maximum number of new tokens to generate.
+
+    Returns:
+        str: The generated response text.
+    """
+    try:
+        # Check if image or text is None
+        if not image or not text:
+            gr.Warning("Please provide an image and a question.")
+
+        # Prepare the inputs
+        text = "answer en " + text
+        inputs = processor(text=text, images=image, return_tensors="pt").to(device)
+
+        # Generate the response
+        with torch.inference_mode():
+            generated_ids = model.generate(
+                **inputs, max_new_tokens=max_new_tokens, do_sample=sampling
+            )
+
+        # Decode the generated response
+        result = processor.batch_decode(generated_ids, skip_special_tokens=True)
+
+        # Log the successful generation of the answer
+        logging.info("Answer generated successfully.")
+
+        # Return the generated response
+        return result[0][len(text) :].lstrip("\n")
+
+    # Handle exceptions that may occur during answer generation
+    except Exception as e:
+        # Custom exception handling
+        raise CustomExceptionHandling(e, sys) from e