Update app.py

app.py

@@ -1,29 +1,32 @@
 import gradio as gr
-from transformers import pipeline
+from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
 
-# Load the model using the pipeline (more stable for Gradio)
-# Blenderbot-400M can sometimes hit memory limits; pipeline is efficient.
-chatbot_pipeline = pipeline("text2text-generation", model="facebook/blenderbot-400M-distill")
+model_name = "facebook/blenderbot-400M-distill"
+
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
 
 def chat_function(message, history):
-    # Gradio history is a list of [user, bot] pairs. 
-    # We turn it into one string for Blenderbot.
-    conversation = ""
-    for user_msg, bot_msg in history:
-        conversation += f"{user_msg} {bot_msg} "
-    
-    conversation += message
-
-    # Generate the response
-    # truncation=True prevents "input too long" errors
-    result = chatbot_pipeline(conversation, max_new_tokens=60, truncation=True)
-    
-    return result[0]['generated_text']
-
-# Define the interface
+
+    history_text = ""
+
+    for user, bot in history:
+        history_text += user + " " + bot + " "
+
+    history_text += message
+
+    inputs = tokenizer(history_text, return_tensors="pt", truncation=True)
+
+    outputs = model.generate(**inputs, max_new_tokens=60)
+
+    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
+
+    return response
+
+
 demo = gr.ChatInterface(
     fn=chat_function,
-    title="Blenderbot Chat",
+    title="BlenderBot Chat",
     description="Ask me anything!"
 )