Update app.py

app.py

@@ -1164,27 +1164,30 @@ def display_historical_data(storage_location: str, url: str):
         logging.error(f"Error fetching historical data for {url}: {e}")
         return pd.DataFrame()
 
-# Function to load the "google/flan-t5-xl" model
-def load_model():
+    def load_model():
     """
     Loads the openLlama model and tokenizer once and returns the pipeline.
     """
     try:
-        model_name = "openlm-research/open_llama_3b_v2"  # Correct indentation here
+        model_name = "openlm-research/open_llama_3b_v2"
         tokenizer = AutoTokenizer.from_pretrained(model_name, use_fast=False, legacy=False)
-        model = AutoModelForCausalLM.from_pretrained(model_name) 
+        model = AutoModelForCausalLM.from_pretrained(model_name)
 
-# Determine the maximum supported length for the model
-max_supported_length = 2048  # You might need to adjust this
+        # This should be inside the try block
+        max_supported_length = 2048
 
-openllama_pipeline = pipeline(
-    "text-generation",  
+        openllama_pipeline = pipeline(
+            "text-generation",
+            model=model,
+            tokenizer=tokenizer,
+            truncation=True,
+            max_length=max_supported_length,
             temperature=0.7,
             top_p=0.95,
             device=0 if torch.cuda.is_available() else -1,
         )
         logging.info("Model loaded successfully.")
-        return pipe
+        return openllama_pipeline  # Return the pipeline
     except Exception as e:
         logging.error(f"Error loading google/flan-t5-xl model: {e}")
         return None