Update app.py

app.py

@@ -1,72 +1,98 @@
 import gradio as gr
-import os
-from huggingface_hub import InferenceClient
+
 import torch
-from transformers import AutoTokenizer, AutoModelForCausalLM
+from transformers import (
+    AutoTokenizer,
+    AutoModelForCausalLM,
+    TextIteratorStreamer,
+    pipeline,
+)
+from threading import Thread
 
-"""
-For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
-"""
-repo_id = "Mikhil-jivus/Llama-32-3B-FineTuned"
 access_token = os.getenv('HF_TOKEN')
-# Load the tokenizer and model from the Hugging Face repository
-tokenizer = AutoTokenizer.from_pretrained(repo_id, trust_remote_code=True, token=access_token)
+
+# The huggingface model id for Finetuned model
+checkpoint = "Mikhil-jivus/Llama-32-3B-FineTuned"
+
+# Download and load model and tokenizer
+tokenizer = AutoTokenizer.from_pretrained(checkpoint, trust_remote_code=True,token=access_token)
+model = AutoModelForCausalLM.from_pretrained(
+    checkpoint, torch_dtype=torch.bfloat16, device_map="auto", trust_remote_code=True,token=access_token
+)
+
+# Text generation pipeline
+phi2 = pipeline(
+    "text-generation",
+    tokenizer=tokenizer,
+    model=model,
+    pad_token_id=tokenizer.eos_token_id,
+    eos_token_id=tokenizer.eos_token_id,
+    device_map="auto",
+)
 
 
-client = InferenceClient(model=repo_id,tokenizer=tokenizer,token = access_token)
+# Function that accepts a prompt and generates text using the phi2 pipeline
+def generate(message, chat_history, max_new_tokens):
+    instruction = "You are a helpful assistant to 'User'. You do not respond as 'User' or pretend to be 'User'. You only respond once as 'Assistant'."
+    final_prompt = f"Instruction: {instruction}\n"
 
+    for sent, received in chat_history:
+        final_prompt += "User: " + sent + "\n"
+        final_prompt += "Assistant: " + received + "\n"
 
-def respond(
-    message,
-    history: list[tuple[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-):
-    messages = [{"role": "system", "content": system_message}]
+    final_prompt += "User: " + message + "\n"
+    final_prompt += "Output:"
 
-    for val in history:
-        if val[0]:
-            messages.append({"role": "user", "content": val[0]})
-        if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
+    # Streamer
+    streamer = TextIteratorStreamer(
+        tokenizer=tokenizer, skip_prompt=True, skip_special_tokens=True, timeout=300.0
+    )
+    thread = Thread(
+        target=phi2,
+        kwargs={
+            "text_inputs": final_prompt,
+            "max_new_tokens": max_new_tokens,
+            "streamer": streamer,
+        },
+    )
+    thread.start()
 
-    messages.append({"role": "user", "content": message})
+    generated_text = ""
+    for word in streamer:
+        generated_text += word
+        response = generated_text.strip()
 
-    response = ""
+        if "User:" in response:
+            response = response.split("User:")[0].strip()
 
-    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
+        if "Assistant:" in response:
+            response = response.split("Assistant:")[1].strip()
 
-        response += token
         yield response
 
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
-demo = gr.ChatInterface(
-    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)",
-        ),
-    ],
-)
 
+# Chat interface with gradio
+with gr.Blocks() as demo:
+    gr.Markdown(
+        """
+  # Jivus AI Chatbot Demo
+  This chatbot was created using Llama 3 billion parameter Transformer model. 
+  """
+    )
+
+    tokens_slider = gr.Slider(
+        8,
+        128,
+        value=21,
+        label="Maximum new tokens",
+        info="A larger `max_new_tokens` parameter value gives you longer text responses but at the cost of a slower response time.",
+    )
+
+    chatbot = gr.ChatInterface(
+        fn=generate,
+        additional_inputs=[tokens_slider],
+        stop_btn=None,
+        examples=[["Who is Leonhard Euler?"]],
+    )
 
-if __name__ == "__main__":
-    demo.launch()
+demo.queue().launch()