Update app.py

app.py

@@ -1,146 +1,122 @@
 import gradio as gr
+import torch
+from transformers import (
+    AutoModelForCausalLM,
+    AutoTokenizer,
+    TextIteratorStreamer,
+    BitsAndBytesConfig,
+)
 import os
-import spaces
-from transformers import GemmaTokenizer, AutoModelForCausalLM
-from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
 from threading import Thread
+import spaces
+import time
 
-# Set an environment variable
-HF_TOKEN = os.environ.get("HF_TOKEN", None)
-
-
-DESCRIPTION = '''
-<div>
-<h1 style="text-align: center;">Meta Llama3 8B</h1>
-<p>This Space demonstrates the instruction-tuned model <a href="https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct"><b>Meta Llama3 8b Chat</b></a>. Meta Llama3 is the new open LLM and comes in two sizes: 8b and 70b. Feel free to play with it, or duplicate to run privately!</p>
-<p>🔎 For more details about the Llama3 release and how to use the model with <code>transformers</code>, take a look <a href="https://huggingface.co/blog/llama3">at our blog post</a>.</p>
-<p>🦕 Looking for an even more powerful model? Check out the <a href="https://huggingface.co/chat/"><b>Hugging Chat</b></a> integration for Meta Llama 3 70b</p>
-</div>
-'''
-
-LICENSE = """
-<p/>
-
----
-Built with Meta Llama 3
-"""
-
-PLACEHOLDER = """
-<div style="padding: 30px; text-align: center; display: flex; flex-direction: column; align-items: center;">
-   <img src="https://ysharma-dummy-chat-app.hf.space/file=/tmp/gradio/8e75e61cc9bab22b7ce3dec85ab0e6db1da5d107/Meta_lockup_positive%20primary_RGB.jpg" style="width: 80%; max-width: 550px; height: auto; opacity: 0.55;  "> 
-   <h1 style="font-size: 28px; margin-bottom: 2px; opacity: 0.55;">Meta llama3</h1>
-   <p style="font-size: 18px; margin-bottom: 2px; opacity: 0.65;">Ask me anything...</p>
-</div>
-"""
-
-
-css = """
-h1 {
-  text-align: center;
-  display: block;
-}
+token = os.environ["HF_TOKEN"]
 
-#duplicate-button {
-  margin: auto;
-  color: white;
-  background: #1565c0;
-  border-radius: 100vh;
-}
-"""
+quantization_config = BitsAndBytesConfig(
+    load_in_4bit=True, bnb_4bit_compute_dtype=torch.float16
+)
 
-# Load the tokenizer and model
-tokenizer = AutoTokenizer.from_pretrained("chheplo/sft_8b_2_llama3")
-model = AutoModelForCausalLM.from_pretrained("chheplo/sft_8b_2_llama3", device_map="auto")  # to("cuda:0") 
+model = AutoModelForCausalLM.from_pretrained(
+    "chheplo/sft_8b_2_llama3", quantization_config=quantization_config, token=token
+)
+tok = AutoTokenizer.from_pretrained("chheplo/sft_8b_2_llama3", token=token)
 terminators = [
-    tokenizer.eos_token_id,
-    tokenizer.convert_tokens_to_ids("<|eot_id|>")
+    tok.eos_token_id,
+    tok.convert_tokens_to_ids("<|eot_id|>")
 ]
 
-@spaces.GPU(duration=120)
-def chat_llama3_8b(message: str, 
-              history: list, 
-              temperature: float, 
-              max_new_tokens: int
-             ) -> str:
-    """
-    Generate a streaming response using the llama3-8b model.
-    Args:
-        message (str): The input message.
-        history (list): The conversation history used by ChatInterface.
-        temperature (float): The temperature for generating the response.
-        max_new_tokens (int): The maximum number of new tokens to generate.
-    Returns:
-        str: The generated response.
-    """
-    conversation = []
-    for user, assistant in history:
-        conversation.extend([{"role": "user", "content": user}, {"role": "assistant", "content": assistant}])
-    conversation.append({"role": "user", "content": message})
+if torch.cuda.is_available():
+    device = torch.device("cuda")
+    print(f"Using GPU: {torch.cuda.get_device_name(device)}")
+else:
+    device = torch.device("cpu")
+    print("Using CPU")
+
+# model = model.to(device)
+# Dispatch Errors
 
-    input_ids = tokenizer.apply_chat_template(conversation, return_tensors="pt").to(model.device)
-    
-    streamer = TextIteratorStreamer(tokenizer, timeout=10.0, skip_prompt=True, skip_special_tokens=True)
 
+@spaces.GPU()
+def chat(message, history, temperature,do_sample, max_tokens):
+    prompt_template = """
+    You are a helpful Agricultural assistant for farmers. You are given the following input. Please complete the response briefly.
+    ## Question:
+    {}
+    
+    ## Response:
+    {}"""
+    start_time = time.time()
+    chat = []
+    # for item in history:
+    #     chat.append({"role": "user", "content": item[0]})
+    #     if item[1] is not None:
+    #         chat.append({"role": "assistant", "content": item[1]})
+    # chat.append({"role": "user", "content": message})
+    # messages = tok.apply_chat_template(chat, tokenize=False, add_generation_prompt=True)
+    
+    model_inputs = tok(prompt_template.format(
+        message, #input
+        "" # response
+    ), return_tensors="pt").to(device)
+    streamer = TextIteratorStreamer(
+        tok, timeout=10.0, skip_prompt=True, skip_special_tokens=True
+    )
     generate_kwargs = dict(
-        input_ids= input_ids,
+        model_inputs,
         streamer=streamer,
-        max_new_tokens=max_new_tokens,
+        max_new_tokens=max_tokens,
         do_sample=True,
         temperature=temperature,
+        repetition_penalty=1.2, 
+        use_cache=False,
         eos_token_id=terminators,
     )
-    # This will enforce greedy generation (do_sample=False) when the temperature is passed 0, avoiding the crash.             
+    
     if temperature == 0:
         generate_kwargs['do_sample'] = False
-        
+    
     t = Thread(target=model.generate, kwargs=generate_kwargs)
     t.start()
 
-    outputs = []
-    for text in streamer:
-        outputs.append(text)
-        #print(outputs)
-        yield "".join(outputs)
-        
-
-# Gradio block
-chatbot=gr.Chatbot(height=450, placeholder=PLACEHOLDER, label='Gradio ChatInterface')
-
-with gr.Blocks(fill_height=True, css=css) as demo:
-    
-    gr.Markdown(DESCRIPTION)
-    gr.DuplicateButton(value="Duplicate Space for private use", elem_id="duplicate-button")
-    gr.ChatInterface(
-        fn=chat_llama3_8b,
-        chatbot=chatbot,
-        fill_height=True,
-        additional_inputs_accordion=gr.Accordion(label="⚙️ Parameters", open=False, render=False),
-        additional_inputs=[
-            gr.Slider(minimum=0,
-                      maximum=1, 
-                      step=0.1,
-                      value=0.95, 
-                      label="Temperature", 
-                      render=False),
-            gr.Slider(minimum=128, 
-                      maximum=4096,
-                      step=1,
-                      value=512, 
-                      label="Max new tokens", 
-                      render=False ),
-            ],
-        examples=[
-            ['How to setup a human base on Mars? Give short answer.'],
-            ['Explain theory of relativity to me like I’m 8 years old.'],
-            ['What is 9,000 * 9,000?'],
-            ['Write a pun-filled happy birthday message to my friend Alex.'],
-            ['Justify why a penguin might make a good king of the jungle.']
-            ],
-        cache_examples=False,
-                     )
-    
-    gr.Markdown(LICENSE)
-    
-if __name__ == "__main__":
-    demo.launch()
-    
+    partial_text = ""
+    first_token_time = None
+    for new_text in streamer:
+        if not first_token_time:
+            first_token_time = time.time() - start_time
+        partial_text += new_text
+        yield partial_text
+
+    total_time = time.time() - start_time
+    tokens = len(tok.tokenize(partial_text))
+    tokens_per_second = tokens / total_time if total_time > 0 else 0
+
+    timing_info = f"\n\nTime taken to first token: {first_token_time:.2f} seconds\nTokens per second: {tokens_per_second:.2f}"
+    yield partial_text +  timing_info
+
+
+demo = gr.ChatInterface(
+    fn=chat,
+    examples=[["I'm a farmer from Odisha, how do I take care of whitefly in my cotton crop?"]],
+    # multimodal=False,
+    additional_inputs_accordion=gr.Accordion(
+        label="⚙️ Parameters", open=False, render=False
+    ),
+    additional_inputs=[
+        gr.Slider(
+            minimum=0, maximum=1, step=0.1, value=0.5, label="Temperature", render=False
+        ),
+        gr.Checkbox(label="Sampling",value=False),
+        gr.Slider(
+            minimum=128,
+            maximum=4096,
+            step=1,
+            value=512,
+            label="Max new tokens",
+            render=False,
+        ),
+    ],
+    stop_btn="Stop Generation",
+    title="Chat With LLMs",
+    description="Now Running [KissanAI/llama3-8b-dhenu-0.1-sft-16bit](https://huggingface.co/KissanAI/llama3-8b-dhenu-0.1-sft-16bit) in 4bit")
+demo.launch()