Comet Atomic En

🚀 Comet-Atomic English Text2Text Generation Model

This project provides a text2text generation solution using the Comet-Atomic English model, which can generate relevant information based on given events.

🚀 Quick Start

Prerequisites

Make sure you have installed the necessary libraries, including transformers and pandas.

Installation

pip install transformers pandas

💻 Usage Examples

Basic Usage

from transformers import T5ForConditionalGeneration
from transformers import T5TokenizerFast as T5Tokenizer
import pandas as pd
model = "svjack/comet-atomic-en"
device = "cpu"
#device = "cuda:0"
tokenizer = T5Tokenizer.from_pretrained(model)
model = T5ForConditionalGeneration.from_pretrained(model).to(device).eval()

NEED_PREFIX = 'What are the necessary preconditions for the next event?'
EFFECT_PREFIX = 'What could happen after the next event?'
INTENT_PREFIX = 'What is the motivation for the next event?'
REACT_PREFIX = 'What are your feelings after the following event?'


event = "X had a big meal."
for prefix in [NEED_PREFIX, EFFECT_PREFIX, INTENT_PREFIX, REACT_PREFIX]:
    prompt = "{}{}".format(prefix, event)
    encode = tokenizer(prompt, return_tensors='pt').to(device)
    answer = model.generate(encode.input_ids,
                           max_length = 128,
        num_beams=2,
        top_p = 0.95,
        top_k = 50,
        repetition_penalty = 2.5,
        length_penalty=1.0,
        early_stopping=True,
                           )[0]
    decoded = tokenizer.decode(answer, skip_special_tokens=True)
    print(prompt, "\n---Answer：", decoded, "----\n")

Output Example

What are the necessary preconditions for the next event?X had a big meal. 
---Answer： X goes shopping at the supermarket ----

What could happen after the next event?X had a big meal. 
---Answer： X gets fat ----

What is the motivation for the next event?X had a big meal. 
---Answer： X wants to eat ----

What are your feelings after the following event?X had a big meal. 
---Answer： X tastes good ----