Sanberta

🚀 RoBERTa trained on Sanskrit (SanBERTa)

SanBERTa is a RoBERTa model trained on Sanskrit, which can be used for various NLP tasks in the Sanskrit language.

🚀 Quick Start

The following sections will introduce the dataset, configuration, training, evaluation, and usage examples of SanBERTa.

✨ Features

• Trained on Sanskrit: Specifically designed for Sanskrit language processing.
• Multiple Usage Scenarios: Can be used for embedding generation and masked prediction tasks.

📦 Installation

No specific installation steps are provided in the original document.

💻 Usage Examples

Basic Usage

For Embeddings

tokenizer = AutoTokenizer.from_pretrained("surajp/SanBERTa")
model = RobertaModel.from_pretrained("surajp/SanBERTa")

op = tokenizer.encode("इयं भाषा न केवलं भारतस्य अपि तु विश्वस्य प्राचीनतमा भाषा इति मन्यते।", return_tensors="pt")
ps = model(op)
ps[0].shape

'''
Output:
--------
torch.Size([1, 47, 768])

For Prediction

from transformers import pipeline

fill_mask = pipeline(
    "fill-mask",
    model="surajp/SanBERTa",
    tokenizer="surajp/SanBERTa"
)

## इयं भाषा न केवलं भारतस्य अपि तु विश्वस्य प्राचीनतमा भाषा इति मन्यते।
fill_mask("इयं भाषा न केवल<mask> भारतस्य अपि तु विश्वस्य प्राचीनतमा भाषा इति मन्यते।")

ps = model(torch.tensor(enc).unsqueeze(1))
print(ps[0].shape)

'''
Output:
--------
[{'score': 0.7516744136810303,
  'sequence': '<s> इयं भाषा न केवलं भारतस्य अपि तु विश्वस्य प्राचीनतमा भाषा इति मन्यते।</s>',
  'token': 280,
  'token_str': 'à¤Ĥ'},
 {'score': 0.06230105459690094,
  'sequence': '<s> इयं भाषा न केवली भारतस्य अपि तु विश्वस्य प्राचीनतमा भाषा इति मन्यते।</s>',
  'token': 289,
  'token_str': 'à¥Ģ'},
 {'score': 0.055410224944353104,
  'sequence': '<s> इयं भाषा न केवला भारतस्य अपि तु विश्वस्य प्राचीनतमा भाषा इति मन्यते।</s>',
  'token': 265,
  'token_str': 'ा'},
  ...]

📚 Documentation

Dataset

• Wikipedia articles (used in iNLTK). It contains an evaluation set.
• Sanskrit scraps from CLTK

Configuration

Training

• Hardware: On TPU
• Task: For language modelling
• Strategy: Iteratively increasing --block_size from 128 to 256 over epochs

Evaluation

📄 License

No license information is provided in the original document.

Citation

@misc{Parmar2020Sanberta,
  author = {Parmar, Suraj},
  title = {SanBERTa - a RoBERTa trained on Sanskrit},
  year = {2020},
  month = {Jun},
  publisher = {Hugging Face Model Hub},
  url = {https://huggingface.co/surajp/SanBERTa}
}

Created by Suraj Parmar/@parmarsuraj99 | LinkedIn

Made with ♥ in India