Update app.py

app.py

@@ -3,10 +3,10 @@ from huggingface_hub import InferenceClient
 from transformers import pipeline
 from typing import List, Tuple  # Importing for type annotations
 
-# Initialize InferenceClient with the correct model
-client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
+# Initialize the BERT model pipeline for the "fill-mask" task
+pipe = pipeline("fill-mask", model="bert-base-uncased")
 
-# Function to handle the response generation using chat completion
+# Function to handle the response generation using BERT
 def respond(
     message: str,
     history: List[Tuple[str, str]],  # Using List and Tuple for type annotation
@@ -29,17 +29,14 @@ def respond(
 
     response = ""
 
-    # Use the client to get chat completion and stream the response
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
-        response += token
-        yield response
+    # Using the BERT pipeline to fill the mask (this is different from the GPT-style completion)
+    # BERT doesn't generate text the same way, so we are simulating a response for this demo
+    result = pipe(f"Hello, how are you today? {message} [MASK]")
+
+    # Collecting the filled-in mask (likely output from BERT's "fill-mask" task)
+    response = result[0]['sequence']
+
+    yield response
 
 
 # Setting up Gradio Interface