app.py

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

# Load the implicit CoT model
implicit_cot_model_name = 'yuntian-deng/implicit-cot-math-mistral7b'
implicit_cot_model = AutoModelForCausalLM.from_pretrained(implicit_cot_model_name, torch_dtype=torch.bfloat16)
tokenizer = AutoTokenizer.from_pretrained(implicit_cot_model_name)

# Constants
MAX_RESULT_TOKENS = 10

@spaces.GPU
def predict_answer(question):
    input_text = ' '.join(question.split()).strip() + ' ' + tokenizer.eos_token
    inputs = tokenizer(input_text, return_tensors='pt').to('cuda' if torch.cuda.is_available() else 'cpu')
    implicit_cot_model.to('cuda' if torch.cuda.is_available() else 'cpu')

    input_ids = inputs['input_ids']
    outputs = implicit_cot_model.generate(input_ids=input_ids,
            max_new_tokens=MAX_RESULT_TOKENS,
            do_sample=False)

    prediction = tokenizer.decode(outputs[0], skip_special_tokens=True)

    return prediction


demo = gr.Interface(
    fn=predict_answer,
    inputs=[
        gr.Textbox(label='Question', value='A set of 7 spoons costs $21. If each spoon would be sold separately, how much would 5 spoons cost?'),
    ],
    outputs=[
        gr.Textbox(label='Implicit CoT Prediction'),
    ],
    title='Solving Grade School Math Problems with Implicit CoT',
    description='This demo showcases Mistral-7B\'s ability to solve grade school math problems without producing intermediate steps, using our stepwise internalization method.',
    article="""
    - [Paper 1: Implicit Chain of Thought Reasoning via Knowledge Distillation](https://arxiv.org/pdf/2311.01460)
    - [Paper 2: 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)
    """,
    clear_btn=None,
    submit_btn="Get Answer!",
    live=False,
    concurrency_limit=1
)
demo.queue(max_size=5).launch()