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

model_name = "Qwen/Qwen2.5-7B-Instruct"

model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype="auto",
    device_map="auto", 
)
tokenizer = AutoTokenizer.from_pretrained(model_name)

@spaces.GPU
def generate(prompt):
    text = tokenizer.apply_chat_template(
        prompt,
        tokenize=False,
        add_generation_prompt=True
    )
    model_inputs = tokenizer([text], return_tensors="pt").to(model.device)
    generated_ids = model.generate(
        **model_inputs,
        max_new_tokens=2048,
    )
    generated_ids = [
        output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
    ]
    response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]

    return response


gr.Interface(
    fn=generate,
    inputs=gr.Text(),
    outputs=gr.Text(),
).launch()