app.py

from huggingface_hub import InferenceClient
import gradio as gr
import os

API_URL = {
    "Mistral" : "https://api-inference.huggingface.co/models/mistralai/Mistral-7B-Instruct-v0.3",
    "Mixtral" : "https://api-inference.huggingface.co/models/mistralai/Mixtral-8x7B-Instruct-v0.1",
    "Mathstral" : "https://api-inference.huggingface.co/models/mistralai/mathstral-7B-v0.1",
}

HF_TOKEN = os.environ['HF_TOKEN']
Hinglish_Prompt = os.environ['Hinglish_Prompt']

mistralClient = InferenceClient(
    API_URL["Mistral"],
    headers = {"Authorization" : f"Bearer {HF_TOKEN}"},
)

mixtralClient = InferenceClient(
    model = API_URL["Mixtral"],
    headers = {"Authorization" : f"Bearer {HF_TOKEN}"},
)

mathstralClient = InferenceClient(
    model = API_URL["Mathstral"],
    headers = {"Authorization" : f"Bearer {HF_TOKEN}"},
)

def format_prompt(message, history, enable_hinglish=False):
  prompt = "<s>"
  # Adding the Hinglish prompt
  if enable_hinglish and not any("You are a Hinglish LLM." in user_prompt for user_prompt, bot_response in history):
      prompt += Hinglish_Prompt
      
  for user_prompt, bot_response in history:
    prompt += f"[INST] {user_prompt} [/INST]"
    prompt += f" {bot_response}</s> "
  prompt += f"[INST] {message} [/INST]"
  return prompt

def generate(prompt, history, temperature=0.9, max_new_tokens=256, top_p=0.95, repetition_penalty=1.0, model = "Mistral", enable_hinglish=False):
    if model == "Mathstral" and enable_hinglish:
        raise gr.Error("Mathstral decoder doesn't support Hinglish")
        enable_hinglish=False

    # Selecting model to be used
    if(model == "Mistral"):
      client = mistralClient
    elif(model == "Mixstral"):
      client = mixtralClient
    elif(model == "Mathstral"):
      client = mixtralClient

    
    temperature = float(temperature) # Generation arguments
    if temperature < 1e-2:
        temperature = 1e-2
        
    top_p = float(top_p)
    
    generate_kwargs = dict(
        temperature=temperature,
        max_new_tokens=max_new_tokens,
        top_p=top_p,
        repetition_penalty=repetition_penalty,
        do_sample=True,
        seed=42,
    )
    
    formatted_prompt = format_prompt(prompt, history, enable_hinglish)
    stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
    output = ""
    for response in stream:
        output += response.token.text
        yield output
    return output

additional_inputs=[
    gr.Slider(
        label="Temperature",
        value=0.9,
        minimum=0.0,
        maximum=1.0,
        step=0.05,
        interactive=True,
        info="Higher values produce more diverse outputs",
    ),
    gr.Slider(
        label="Max new tokens",
        value=256,
        minimum=0,
        maximum=1048,
        step=64,
        interactive=True,
        info="The maximum numbers of new tokens",
    ),
    gr.Slider(
        label="Top-p (nucleus sampling)",
        value=0.90,
        minimum=0.0,
        maximum=1,
        step=0.05,
        interactive=True,
        info="Higher values sample more low-probability tokens",
    ),
    gr.Slider(
        label="Repetition penalty",
        value=1.2,
        minimum=1.0,
        maximum=2.0,
        step=0.05,
        interactive=True,
        info="Penalize repeated tokens",
    ),
        gr.Dropdown(
        choices = ["Mistral","Mixtral", "Mathstral"],
        value = "Mistral",
        label = "Model to be used",
        interactive=True,
        info = "Mistral : for general-purpose conversations, "+ 
               "Mixtral : for faster and accurate conversations, "+ 
               "Mathstral : for mathematical and scientific reasoning"
    ),
    gr.Checkbox(
        label="Hinglish",
        value=False,
        interactive=True,
        info="Enables the MistralTalk to talk in Hinglish (Combination of Hindi and English)",
    )
]

css = """
  #mkd {
    height: 500px;
    overflow: auto;
    border: 1px solid #ccc;
  }
"""

with gr.Blocks(css=css) as demo:
    gr.HTML("<h1><center>MistralTalk🗣️</center><h1>")
    gr.HTML("<h3><center>In this demo, you can chat with <a href='https://huggingface.co/mistralai'>Mistral AI advanced models</a> 💬</center><h3>")
    gr.HTML("<h3><center>Learn more about the model <a href='https://huggingface.co/docs/transformers/main/model_doc/mistral'>here</a>. 📚</center><h3>")
    gr.ChatInterface(
        generate,
        additional_inputs=additional_inputs,
        theme = gr.themes.Soft(),
        examples=[["What is the secret to life?"], ["How the universe works?"],["What is quantum mechanics?"],["Do you belive in after life?"],["Albert likes to surf every week. Each surfing session lasts for 4 hours and costs $20 per hour. How much would Albert spend in 5 weeks?"]]
    )

demo.queue(max_size=100).launch(debug=True)