metadata

license: apache-2.0
datasets:
  - asset
  - wi_locness
  - GEM/wiki_auto_asset_turk
  - discofuse
  - zaemyung/IteraTeR_plus
  - jfleg
language:
  - en
metrics:
  - sari
  - bleu
  - accuracy

Model Card for CoEdIT-xl-composite

This model was obtained by fine-tuning the corresponding google/flan-t5-xl model on the CoEdIT-Composite dataset. Details of the dataset can be found in our paper and repository.

Paper: CoEdIT: Text Editing by Task-Specific Instruction Tuning

Authors: Vipul Raheja, Dhruv Kumar, Ryan Koo, Dongyeop Kang

Model Details

Model Description

Language(s) (NLP): English
Finetuned from model: google/flan-t5-xl

Model Sources [optional]

Repository: https://github.com/vipulraheja/coedit
Paper [optional]: [More Information Needed]

How to use

We make available the models presented in our paper.

Model	Number of parameters
CoEdIT-large	770M
CoEdIT-xl	3B
CoEdIT-xxl	11B

Uses

Text Revision Task

Given an edit instruction and an original text, our model can generate the edited version of the text.

This model can also perform edits on composite instructions, as shown below:

Usage

from transformers import AutoTokenizer, T5ForConditionalGeneration

tokenizer = AutoTokenizer.from_pretrained("grammarly/coedit-xl-composite")
model = T5ForConditionalGeneration.from_pretrained("grammarly/coedit-xl-composite")
input_text = 'Fix grammatical errors in this sentence and make it simpler: New kinds of vehicles will be invented with new technology than today.'
input_ids = tokenizer(input_text, return_tensors="pt").input_ids
outputs = model.generate(input_ids, max_length=256)
edited_text = tokenizer.decode(outputs[0], skip_special_tokens=True)[0]

Software

https://github.com/vipulraheja/coedit

Citation

BibTeX:

[More Information Needed]

APA:

[More Information Needed]