Deberta V3 Small Finetuned Cola

🚀 DeBERTa-v3-small fine-tuned on CoLA

This model is a fine - tuned version of microsoft/deberta-v3-small on the GLUE COLA dataset. It can effectively handle text - classification tasks on the CoLA dataset, achieving good performance in evaluation.

🚀 Quick Start

This section is not provided in the original document, so it is skipped.

✨ Features

• Enhanced Architecture: DeBERTa improves the BERT and RoBERTa models using disentangled attention and enhanced mask decoder.
• New Pre - training Objective: In DeBERTa V3, the MLM objective is replaced with the RTD (Replaced Token Detection) objective introduced by ELECTRA for pre - training.
• Good Performance: The model significantly improves the performance in downstream tasks compared to DeBERTa - V2.

📚 Documentation

Model description

DeBERTa improves the BERT and RoBERTa models using disentangled attention and enhanced mask decoder. With those two improvements, DeBERTa out perform RoBERTa on a majority of NLU tasks with 80GB training data.

Please check the official repository for more details and updates.

In DeBERTa V3, we replaced the MLM objective with the RTD(Replaced Token Detection) objective introduced by ELECTRA for pre - training, as well as some innovations to be introduced in our upcoming paper. Compared to DeBERTa - V2, our V3 version significantly improves the model performance in downstream tasks. You can find a simple introduction about the model from the appendix A11 in our original paper, but we will provide more details in a separate write - up.

The DeBERTa V3 small model comes with 6 layers and a hidden size of 768. Its total parameter number is 143M since we use a vocabulary containing 128K tokens which introduce 98M parameters in the Embedding layer. This model was trained using the 160GB data as DeBERTa V2.

Intended uses & limitations

More information needed

Training and evaluation data

The Corpus of Linguistic Acceptability (CoLA) in its full form consists of 10657 sentences from 23 linguistics publications, expertly annotated for acceptability (grammaticality) by their original authors. The public version provided here contains 9594 sentences belonging to training and development sets, and excludes 1063 sentences belonging to a held out test set.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 3e - 05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon = 1e - 08
• lr_scheduler_type: linear
• num_epochs: 5.0

Training results

Framework versions

• Transformers 4.13.0.dev0
• Pytorch 1.10.0+cu111
• Datasets 1.16.1
• Tokenizers 0.10.3

📄 License

This model is released under the MIT license.