3

app.py

@@ -1,15 +1,22 @@
 import gradio as gr
 from transformers import pipeline
 
-# Load a more suitable model for conversational responses
-model_name = "gpt2"  # You might want to try 'gpt-neo' or 'gpt-3.5-turbo' if available
+# Load the model
+model_name = "gpt2"
 generator = pipeline("text-generation", model=model_name)
 
 # Inference function
 def generate_response(prompt):
-    # Generate text with a more structured approach
-    response = generator(prompt, max_length=100, num_return_sequences=1)[0]['generated_text']
-    return response.strip()  # Clean up any leading/trailing whitespace
+    # Generate text with specific parameters
+    response = generator(
+        prompt,
+        max_length=150,  # Increase max length for more comprehensive responses
+        num_return_sequences=1,
+        temperature=0.7,  # Lower for more deterministic responses
+        top_k=50,         # Consider the top 50 tokens for diversity
+        top_p=0.95        # Cumulative probability for diversity
+    )
+    return response[0]['generated_text'].strip()  # Clean up the output
 
 # Gradio interface
 interface = gr.Interface(
@@ -17,7 +24,7 @@ interface = gr.Interface(
     inputs="text", 
     outputs="text", 
     title="Conversational LLM",
-    description="Enter a message to receive a relevant response."
+    description="Enter a prompt to generate a relevant and coherent response."
 )
 
 # Launch the interface