Update app.py

app.py

@@ -47,7 +47,7 @@ from transformers import AutoModelForCausalLM, AutoTokenizer
 # Load the model and tokenizer
 @st.cache_resource
 def load_model_and_tokenizer():
-    model_name = "EleutherAI/gpt-neo-2.7B"  # Model suited for longer responses
+    model_name = "microsoft/DialoGPT-medium"  # Replace with your chosen model
     tokenizer = AutoTokenizer.from_pretrained(model_name)
     model = AutoModelForCausalLM.from_pretrained(model_name)
     return tokenizer, model
@@ -55,43 +55,46 @@ def load_model_and_tokenizer():
 tokenizer, model = load_model_and_tokenizer()
 
 # Streamlit App
-st.title("General Chatbot with Detailed Responses")
+st.title("General Chatbot")
 st.write("A chatbot powered by an open-source model from Hugging Face.")
 
-# Initialize the conversation
+# Initialize the conversation history
 if "conversation_history" not in st.session_state:
     st.session_state["conversation_history"] = []
 
-# Input for user query
+# Input box for user query
 user_input = st.text_input("You:", placeholder="Ask me anything...", key="user_input")
 
-# Slider for response length
-max_length = st.slider("Set maximum response length:", min_value=100, max_value=1000, step=50, value=300)
-
 if st.button("Send") and user_input:
-    # Add user query to the conversation
+    # Append user input to history
     st.session_state["conversation_history"].append({"role": "user", "content": user_input})
-    
-    # Build prompt for the model
-    prompt = f"{user_input} Please explain in detail."
-    input_ids = tokenizer.encode(prompt + tokenizer.eos_token, return_tensors="pt")
-    
-    # Generate response
+
+    # Prepare the input for the model
+    conversation_context = ""
+    for message in st.session_state["conversation_history"]:
+        if message["role"] == "user":
+            conversation_context += f"User: {message['content']}\n"
+        elif message["role"] == "assistant":
+            conversation_context += f"Bot: {message['content']}\n"
+
+    input_ids = tokenizer.encode(conversation_context + "Bot:", return_tensors="pt")
+
+    # Generate the response with adjusted parameters
     chat_history_ids = model.generate(
         input_ids,
-        max_length=max_length,
-        temperature=0.9,
-        top_p=0.9,
-        repetition_penalty=1.2,
-        pad_token_id=tokenizer.eos_token_id,
-        early_stopping=False
+        max_length=500,  # Increase maximum length for longer responses
+        num_return_sequences=1,
+        temperature=0.7,  # Adjust for creativity (lower is more focused, higher is more diverse)
+        top_p=0.9,  # Use nucleus sampling for diversity
+        top_k=50,  # Limit to top-k tokens for more controlled output
+        pad_token_id=tokenizer.eos_token_id
     )
-    response = tokenizer.decode(chat_history_ids[:, input_ids.shape[-1]:][0], skip_special_tokens=True)
 
-    # Add bot response to the conversation
+    # Decode the response and add it to history
+    response = tokenizer.decode(chat_history_ids[:, input_ids.shape[-1]:][0], skip_special_tokens=True)
     st.session_state["conversation_history"].append({"role": "assistant", "content": response})
 
-# Display conversation history
+# Display the conversation
 for message in st.session_state["conversation_history"]:
     if message["role"] == "user":
         st.write(f"**You:** {message['content']}")