Bleurt Tiny 512

🚀 Model Card for bleurt-tiny-512

This is a Pytorch version of the original BLEURT models from an ACL paper, which can be used for text classification tasks.

🚀 Quick Start

Use the code below to get started with the model.

from transformers import AutoModelForSequenceClassification, AutoTokenizer
import torch

tokenizer = AutoTokenizer.from_pretrained("Elron/bleurt-tiny-512")
model = AutoModelForSequenceClassification.from_pretrained("Elron/bleurt-tiny-512")
model.eval()

references = ["hello world", "hello world"]
candidates = ["hi universe", "bye world"]

with torch.no_grad():
  scores = model(**tokenizer(references, candidates, return_tensors='pt'))[0].squeeze()

print(scores) # tensor([-0.9414, -0.5678])

✨ Features

• This is a Pytorch version of the original BLEURT models from an ACL paper.
• It can be used for the task of text classification.

📚 Documentation

Model Details

• Developed by: Elron Bandel, Thibault Sellam, Dipanjan Das and Ankur P. Parikh of Google Research
• Shared by [Optional]: Elron Bandel
• Model type: Text Classification
• Language(s) (NLP): More information needed
• License: More information needed
• Parent Model: BERT
• Resources for more information:
  • GitHub Repo
  • Associated Paper
  • Blog Post

Uses

Direct Use

This model can be used for the task of text classification.

Out-of-Scope Use

The model should not be used to intentionally create hostile or alienating environments for people.

Bias, Risks, and Limitations

Significant research has explored bias and fairness issues with language models (see, e.g., Sheng et al. (2021) and Bender et al. (2021)). Predictions generated by the model may include disturbing and harmful stereotypes across protected classes; identity characteristics; and sensitive, social, and occupational groups.

⚠️ Important Note

Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.

Training Details

Training Data

The model authors note in the associated paper:

We use years 2017 to 2019 of the WMT Metrics Shared Task, to-English language pairs. For each year, we used the of- ficial WMT test set, which include several thou- sand pairs of sentences with human ratings from the news domain. The training sets contain 5,360, 9,492, and 147,691 records for each year.

Testing Data

The test sets for years 2018 and 2019 [of the WMT Metrics Shared Task, to-English language pairs.] are noisier.

Environmental Impact

Carbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).

Citation

BibTeX:

@inproceedings{sellam2020bleurt,
  title = {BLEURT: Learning Robust Metrics for Text Generation},
  author = {Thibault Sellam and Dipanjan Das and Ankur P Parikh},
  year = {2020},
  booktitle = {Proceedings of ACL}
}

Model Card Authors [optional]

Elron Bandel in collaboration with Ezi Ozoani and the Hugging Face team