app.py

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
import os
import gradio as gr

model_name = "scb10x/llama-3-typhoon-v1.5x-70b-instruct-awq"
token = os.getenv("HF_TOKEN")

# Remove these lines
# device = torch.device("cuda")
# torch.cuda.set_device(0)

tokenizer = AutoTokenizer.from_pretrained(model_name, token=token)
model = AutoModelForCausalLM.from_pretrained(model_name, token=token)

def generate_text(prompt):
    inputs = tokenizer(prompt, return_tensors="pt")
    outputs = model.generate(inputs.input_ids, max_length=50)
    return tokenizer.decode(outputs[0], skip_special_tokens=True)

gr.Interface(fn=generate_text, inputs="text", outputs="text").launch()