vagmi/grammar-t5

This is my first model for grammar error correction. It uses the jfleg dataset and is built on t5-base. It is trained only for 3 epochs so the output isn't that great.

Usage

You can use this model with the standard transformers library. This model should be small enough to run on the CPU.

$ pip install transformers torch sentencepiece

Once you have the dependencies setup, you should be able to run this model.

from transformers import AutoModelForSeq2SeqLM, AutoTokenizer

model_name = 'vagmi/grammar-t5'

model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
tokenizer = AutoTokenizer.from_pretrained(model_name)

text = 'fix grammar: I am work with machine to write gooder english.'
inputs = tokenizer(text, return_tensors='pt')
outputs = model.generate(inputs['input_ids'], num_beams=2, max_length=512, early_stopping=True)
fixed = tokenizer.decode(outputs[0], skip_special_tokens=True)
# I am working with machine to write better english.