TaskToSubtask

Overview

This repository contains a Hugging Face NLP model designed to break down long task instructions into smaller, actionable subtasks. The output is a single string where each subtask is separated by a comma.

Installation

To use this model, you'll need to install the transformers library from Hugging Face. You can install it using pip:

pip install sentencepiece
pip install transformers
pip install accelerate -U

Usage

To use the Subtask Generator, you can use the following Python code snippet:

import torch
from transformers import T5Tokenizer, T5ForConditionalGeneration, TrainingArguments, Trainer
model = T5ForConditionalGeneration.from_pretrained("Jwizzed/TaskToSubtask").to('cpu')
tokenizer = T5Tokenizer.from_pretrained("t5-small")

def infer(input_text, model, tokenizer):
    model.eval()
    input_tensor = tokenizer.encode(input_text, return_tensors="pt").to('cpu')
    with torch.no_grad():
        output = model.generate(input_tensor, max_length=1024)
    decoded_output = tokenizer.decode(output[0], skip_special_tokens=True)
    decoded_output = decoded_output.replace(" and ", ", ").strip()
    return decoded_output


sample_text = """
Dear Students
Please watch all class materials for our topic before you answer the following questions.
1. Give your own definition of mental health.
2. What does a person with good mental health look like?
3. How do you deal with your stress?
4. Out of many suggestions to help improve your mental health, which strategies you think most helpful to you?

Please type your answer. Don’t forget to put your name and ID on the right top cornor. Written in full sentences is preferable than mind map

Plagiarism/ copying your friends’ answers will be penalized.
"""

print(infer(sample_text, model.to('cpu'), tokenizer))