app.py

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

model_name = 'yuntian-deng/gpt2-implicit-cot-multiplication'
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)

def preprocess(num):
    num = str(num).strip().replace(' ', '')
    reversed_num = ' '.join(num[::-1])
    return reversed_num

def postprocess(raw_output):
    prediction = raw_output.replace(' ', '')[::-1]
    return prediction

@spaces.GPU
def predict_product(num1, num2):
    input_text = f'{preprocess(num1)} * {preprocess(num2)} ='
    inputs = tokenizer(input_text, return_tensors='pt').to('cuda' if torch.cuda.is_available() else 'cpu')
    model.to('cuda' if torch.cuda.is_available() else 'cpu')
    outputs = model.generate(**inputs, max_new_tokens=40)
    output = outputs[0][inputs['input_ids'].shape[-1]:]
    raw_output = tokenizer.decode(output, skip_special_tokens=True)
    prediction = postprocess(raw_output)
    return input_text, raw_output, prediction

demo = gr.Interface(
    fn=predict_product,
    inputs=[gr.Number(label='First Number (up to 9 digits)'), gr.Number(label='Second Number (up to 9 digits)')],
    outputs=[
        gr.Textbox(label='Raw Input to GPT-2'),
        gr.Textbox(label='Raw Output from GPT-2'),
        gr.Textbox(label='Predicted Product')
    ],
    title='GPT-2 Multiplication Predictor',
    description='Enter two numbers up to 9 digits each and get the predicted product.',
    article="""
    ### Additional Resources
    - [Paper: From Explicit CoT to Implicit CoT: Learning to Internalize CoT Step by Step](https://arxiv.org/pdf/2405.14838)
    - [Code Repository](https://github.com/da03/Internalize_CoT_Step_by_Step)
    - [Tweet Announcement](https://twitter.com/yuntiandeng/status/1795854740879774036)
    """
)

demo.launch()