Update app.py

app.py

@@ -1,4 +1,5 @@
 import gradio as gr
+from huggingface_hub import InferenceClient
 from transformers import AutoTokenizer, AutoModelForCausalLM
 import torch
 import os
@@ -6,41 +7,84 @@ import os
 # Replace 'your_huggingface_token' with your actual Hugging Face access token
 access_token = os.getenv('token')
 
-# Load the tokenizer and model
+# Initialize the tokenizer and model with the Hugging Face access token
 tokenizer = AutoTokenizer.from_pretrained("google/gemma-2b-it", use_auth_token=access_token)
 model = AutoModelForCausalLM.from_pretrained(
     "google/gemma-2b-it",
     torch_dtype=torch.bfloat16,
-    device_map="auto",
-    use_auth_token=access_token# Automatically map to GPU if available
+    use_auth_token=access_token
 )
+model.eval()  # Set the model to evaluation mode
 
-# Define generation function
-def generate_text(prompt, max_tokens=200, temperature=0.7, top_p=0.9):
-    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
-    outputs = model.generate(
-        **inputs,
-        max_new_tokens=max_tokens,
-        do_sample=True,
+# Initialize the inference client (if needed for other API-based tasks)
+client = InferenceClient(provider="together",token=access_token)
+
+def conversation_predict(input_text):
+    """Generate a response for single-turn input using the model."""
+    # Tokenize the input text
+    input_ids = tokenizer(input_text, return_tensors="pt").input_ids
+
+    # Generate a response with the model
+    outputs = model.generate(input_ids, max_new_tokens=2048)
+
+    # Decode and return the generated response
+    return tokenizer.decode(outputs[0], skip_special_tokens=True)
+
+def respond(
+    message: str,
+    history: list[tuple[str, str]],
+    system_message: str,
+    max_tokens: int,
+    temperature: float,
+    top_p: float,
+):
+    """Generate a response for a multi-turn chat conversation."""
+    # Prepare the messages in the correct format for the API
+    messages = [{"role": "system", "content": system_message}]
+
+    for user_input, assistant_reply in history:
+        if user_input:
+            messages.append({"role": "user", "content": user_input})
+        if assistant_reply:
+            messages.append({"role": "assistant", "content": assistant_reply})
+
+    messages.append({"role": "user", "content": message})
+
+    response = ""
+
+    # Stream response tokens from the chat completion API
+    for message_chunk in client.chat_completion(
+        model = "google/gemma-2b-it",
+        messages=messages,
+        max_tokens=max_tokens,
+        stream=True,
         temperature=temperature,
         top_p=top_p,
-        pad_token_id=tokenizer.eos_token_id
-    )
-    return tokenizer.decode(outputs[0], skip_special_tokens=True)
+    ):
+        token = message_chunk["choices"][0]["delta"].get("content", "")
+        response += token
+        yield response
 
-# Define Gradio interface
-interface = gr.Interface(
-    fn=generate_text,
-    inputs=[
-        gr.Textbox(label="Prompt", placeholder="Enter your prompt here...", lines=2),
-        gr.Slider(50, 512, value=200, label="Max Tokens"),
-        gr.Slider(0.1, 1.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(0.1, 1.0, value=0.9, step=0.05, label="Top-p (nucleus sampling)")
+# Create a Gradio ChatInterface demo
+demo = gr.ChatInterface(
+    fn=respond,
+    additional_inputs=[
+        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
+        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
+        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
+        gr.Slider(
+            minimum=0.1,
+            maximum=1.0,
+            value=0.95,
+            step=0.05,
+            label="Top-p (nucleus sampling)",
+        ),
     ],
-    outputs=gr.Textbox(label="Generated Text"),
-    title="Gemma-2B Text Generator",
-    description="Enter a prompt and let Google's Gemma-2B-IT model generate a response."
 )
 
-# Launch the app
-interface.launch()
+if __name__ == "__main__":
+    demo.launch(share=True)
+
+
+
+do not stream the output