app.py

import gradio as gr
import spaces
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM
import subprocess

# Install flash-attn
subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)

# Load the model and tokenizer
model_name = "akjindal53244/Llama-3.1-Storm-8B"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype=torch.bfloat16,
    use_flash_attention_2=True,
    device_map="auto"
)

@spaces.GPU(duration=120)
def generate_text(prompt, max_length, temperature):
    messages = [
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": prompt}
    ]
    formatted_prompt = tokenizer.apply_chat_template(messages, add_generation_prompt=True, tokenize=False)
    
    inputs = tokenizer(formatted_prompt, return_tensors="pt").to("cuda")
    
    outputs = model.generate(
        **inputs,
        max_new_tokens=max_length,
        do_sample=True,
        temperature=temperature,
        top_k=100,
        top_p=0.95,
    )
    
    return tokenizer.decode(outputs[0][inputs['input_ids'].shape[1]:], skip_special_tokens=True)

iface = gr.Interface(
    fn=generate_text,
    inputs=[
        gr.Textbox(lines=5, label="Prompt"),
        gr.Slider(minimum=1, maximum=500, value=128, step=1, label="Max Length"),
        gr.Slider(minimum=0.1, maximum=2.0, value=0.7, step=0.1, label="Temperature"),
    ],
    outputs=gr.Textbox(lines=10, label="Generated Text"),
    title="Llama-3.1-Storm-8B Text Generation",
    description="Enter a prompt to generate text using the Llama-3.1-Storm-8B model.",
)

iface.launch()