Update src/streamlit_app.py

src/streamlit_app.py

@@ -90,17 +90,29 @@ def main():
         data = nutritional_info.get(true_label)
 
         if data:
+            base_description = (
+                f"A typical {true_label} serving ({data['serving']}) contains about {data['calories']}, "
+                f"with {data['protein']} protein, {data['carbs']} carbs, and {data['fat']} fat. "
+                f"Made from {data['ingredients']} and usually {data['method']}. "
+                f"Try {data['substitute']} as a healthier swap."
+            )
             prompt = (
-                f"A typical {true_label} serving ({data['serving']}) contains about {data['calories']}, with "
-                f"{data['protein']} protein, {data['carbs']} carbs, and {data['fat']} fat. Made from "
-                f"{data['ingredients']} and usually {data['method']}. Try {data['substitute']} as a healthier swap.\n\n"
-                f"Paraphrase this nutritional info without changing facts:"
+                f"Paraphrase this nutritional description into a friendly, conversational tone without changing any facts:\n\n"
+                f"{base_description}"
             )
         else:
             prompt = f"Give the typical calories, macros, and nutrition facts for {label}. Provide realistic values even if estimated."
 
         inputs = tokenizer_t5(prompt, return_tensors="pt", truncation=True).to(device)
-        output_ids = model_t5.generate(inputs["input_ids"], max_new_tokens=100, do_sample=True, temperature=0.8, top_p=0.9)
+        output_ids = model_t5.generate(
+            inputs["input_ids"],
+            max_new_tokens=100,
+            do_sample=True,
+            temperature=1.0,
+            top_p=0.95,
+            top_k=40,
+            repetition_penalty=1.1
+        )
         response = tokenizer_t5.decode(output_ids[0], skip_special_tokens=True)
 
         st.subheader("🧾 Nutrition Overview")