app.py

import gradio as gr
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
import spaces
# import pythonexample



pythonexample = """produce a generative ai gradio demo using mistral instruct with the following prompt "i am a helpful assistant that always mentions bannanachicken" for a simple text to text task
"""

title = """# 🙋🏻‍♂️Welcome to Tonic's🪨Granite Code ! """
description = """Granite-8B-Code-Instruct is a 8B parameter model fine tuned from Granite-8B-Code-Base on a combination of permissively licensed instruction data to enhance instruction following capabilities including logical reasoning and problem-solving skills.
### Join us : 
TeamTonic is always making cool demos! Join our active builder's community on Discord: [Discord](https://discord.gg/GWpVpekp) On Huggingface: [TeamTonic](https://huggingface.co/TeamTonic) & [MultiTransformer](https://huggingface.co/MultiTransformer) On Github: [Polytonic](https://github.com/tonic-ai) & contribute to [multitonic](https://github.com/multitonic/multitonic)

### How To Use : 
Add a new line to the example and at the end of your prompts 🚀
"""
# Define the device and model path
device = "cuda" if torch.cuda.is_available() else "cpu"
model_path = "ibm-granite/granite-8b-code-instruct"

# Load the tokenizer and model
tokenizer = AutoTokenizer.from_pretrained(model_path)
model = AutoModelForCausalLM.from_pretrained(model_path)
model.to(device)
model.eval()


# Function to generate code

@spaces.GPU
def generate_code(prompt, max_length):
    # Prepare the input chat format
    chat = [
        { "role": "user", "content": prompt }
    ]
    chat = tokenizer.apply_chat_template(chat, tokenize=False, add_generation_prompt=True)
    # Tokenize the input text
    input_tokens = tokenizer(chat, return_tensors="pt")
    # Transfer tokenized inputs to the device (GPU)
    for i in input_tokens:
        input_tokens[i] = input_tokens[i].to("cuda")
    # Generate output tokens
    output_tokens = model.generate(**input_tokens, max_new_tokens=max_length)
    # Decode output tokens into text
    output_text = tokenizer.batch_decode(output_tokens, skip_special_tokens=True)
    # Return the generated code
    return output_text[0]


# Define Gradio Blocks

def gradio_interface():

    with gr.Blocks() as interface:

        gr.Markdown(title)
        gr.Markdown(description)  

        # Create input and output components
        prompt_input = gr.Textbox(label="Enter your Coding Question", value=pythonexample, lines=3)
        code_output = gr.Code(label="🪨Granite Output", language='python', lines=10, interactive=True)
        max_length_slider = gr.Slider(minimum=1, maximum=2000, value=1000, label="Max Token Length")

        # Create a button to trigger code generation
        generate_button = gr.Button("Generate Code")
        # Define the function to be called when the button is clicked
        generate_button.click(generate_code, inputs=[prompt_input, max_length_slider], outputs=code_output)

    return interface


if __name__ == "__main__":
    # Create and launch the Gradio interface
    interface = gradio_interface()
    interface.launch()