Update app.py

app.py

@@ -20,8 +20,19 @@ def response(user_question, table_data):
 
     queries = [user_question]
 
-    encoding = tokenizer(table=table_data, query=queries, padding=True, return_tensors="pt",truncation=True)
-    outputs = model.generate(**encoding)
+    encoding = tokenizer(table=table_data, query=queries, padding=True, return_tensors="pt", truncation=True)
+
+    # Experiment with generation parameters
+    outputs = model.generate(
+        **encoding,
+        num_beams=5,            # Beam search to generate more diverse responses
+        top_k=50,               # Top-k sampling for diversity
+        top_p=0.95,             # Nucleus sampling
+        temperature=0.7,        # Temperature scaling (if supported by the model)
+        max_length=50,          # Limit the length of the generated response
+        early_stopping=True     # Stop generation when an end token is generated
+    )
+
     ans = tokenizer.batch_decode(outputs, skip_special_tokens=True)
 
     query_result = {
@@ -33,6 +44,7 @@ def response(user_question, table_data):
 
     return query_result
 
+
 # Streamlit interface
 st.markdown("""
 <div style='display: flex; align-items: center;'>