README.md

---
license: mit
language:
- en
metrics:
- precision
- recall
- f1
- accuracy
new_version: v1.0
datasets:
- BookCorpus
- Wikipedia
tags:
- BERT
- MNLI
- NLI
- transformer
- pre-training
- NLP
- MIT-NLP-v1
base_model:
- google/bert-base-uncased
library_name: transformers
---

[![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](https://opensource.org/licenses/MIT)
[![Model Size](https://img.shields.io/badge/Size-~20MB-blue)](#)
[![Type](https://img.shields.io/badge/Type-Compact%20NLP-lightblue)](#)
[![Performance](https://img.shields.io/badge/Recommended%20For-Lightweight%20Capacity-red)](#)

# Model Card for boltuix/bert-tinyplus

The `boltuix/bert-tinyplus` model is an ultra-compact BERT variant designed for natural language processing tasks requiring lightweight performance with slightly better capacity than smaller models like `boltuix/bert-mini`. Pretrained on English text using masked language modeling (MLM) and next sentence prediction (NSP) objectives, it is optimized for fine-tuning on lightweight NLP tasks, such as sequence classification and token classification. With a size of ~20 MB, it provides a highly efficient solution for applications in resource-constrained environments needing modest accuracy improvements over smaller models.

## Model Details

### Model Description

The `boltuix/bert-tinyplus` model is a PyTorch-based transformer model derived from TensorFlow checkpoints in the Google BERT repository. It builds on research from *On the Importance of Pre-training Compact Models* ([arXiv](https://arxiv.org/abs/1908.08962)) and *Generalization in NLI: Ways (Not) To Go Beyond Simple Heuristics* ([arXiv](https://arxiv.org/abs/1908.08962)). Ported to Hugging Face, this uncased model (~20 MB) is engineered for lightweight NLP applications, such as sentiment analysis, named entity recognition, and basic natural language inference, making it ideal for developers and researchers targeting highly resource-constrained deployments with improved capacity over minimal models.

- **Developed by:** BoltUIX
- **Funded by:** BoltUIX Research Fund
- **Shared by:** Hugging Face
- **Model type:** Transformer (BERT)
- **Language(s) (NLP):** English (`en`)
- **License:** MIT
- **Finetuned from model:** google-bert/bert-base-uncased

### Model Sources

- **Repository:** [Hugging Face Model Hub](https://huggingface.co/boltuix/bert-tinyplus)
- **Paper:** [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](http://arxiv.org/abs/1810.04805)
- **Demo:** [Hugging Face Spaces Demo](https://huggingface.co/spaces/boltuix/bert-tinyplus-demo)

## Model Variants

BoltUIX offers a range of BERT-based models tailored to different performance and resource requirements. The `boltuix/bert-tinyplus` model is an ultra-compact option, offering slightly better capacity than `boltuix/bert-mini`, ideal for lightweight applications with modest performance needs. Below is a summary of available models:

| Tier       | Model ID                | Size (MB) | Notes                                              |
|------------|-------------------------|-----------|----------------------------------------------------|
| Micro      | boltuix/bert-micro      | ~15 MB    | Smallest, blazing-fast, moderate accuracy          |
| Mini       | boltuix/bert-mini       | ~17 MB    | Ultra-compact, fast, slightly better accuracy      |
| Tinyplus   | boltuix/bert-tinyplus   | ~20 MB    | Slightly bigger, better capacity                  |
| Small      | boltuix/bert-small      | ~45 MB    | Good compact/accuracy balance                     |
| Mid        | boltuix/bert-mid        | ~50 MB    | Well-rounded mid-tier performance                 |
| Medium     | boltuix/bert-medium     | ~160 MB   | Strong general-purpose model                      |
| Large      | boltuix/bert-large      | ~365 MB   | Top performer below full-BERT                     |
| Pro        | boltuix/bert-pro        | ~420 MB   | Use only if max accuracy is mandatory             |
| Mobile     | boltuix/bert-mobile     | ~140 MB   | Mobile-optimized; quantize to ~25 MB with no major loss |

For more details on each variant, visit the [BoltUIX Model Hub](https://huggingface.co/boltuix).

## Uses

### Direct Use

The model can be used directly for masked language modeling or next sentence prediction tasks, such as predicting missing words in sentences or determining sentence coherence, delivering modest accuracy in these core tasks.

### Downstream Use

The model is designed for fine-tuning on lightweight downstream NLP tasks, including:
- Sequence classification (e.g., basic sentiment analysis, intent detection)
- Token classification (e.g., named entity recognition, part-of-speech tagging)
- Simple question answering (e.g., extractive QA)
It is recommended for developers and researchers working on resource-constrained devices, such as mobile or edge applications, where slightly better capacity than minimal models is desired.

### Out-of-Scope Use

The model is not suitable for:
- Text generation tasks (use generative models like GPT-3 instead).
- Non-English language tasks without significant fine-tuning.
- High-performance applications requiring robust accuracy (use `boltuix/bert-mid`, `boltuix/bert-large`, or `boltuix/bert-pro` instead).

## Bias, Risks, and Limitations

The model may inherit biases from its training data (BookCorpus and English Wikipedia), potentially reinforcing stereotypes, such as gender or occupational biases. For example:
```python
from transformers import pipeline
unmasker = pipeline('fill-mask', model='boltuix/bert-tinyplus')
unmasker("The man worked as a [MASK].")
```
**Output**:
```json
[
  {'sequence': '[CLS] the man worked as a engineer. [SEP]', 'token_str': 'engineer'},
  {'sequence': '[CLS] the man worked as a doctor. [SEP]', 'token_str': 'doctor'},
  ...
]
```
```python
unmasker("The woman worked as a [MASK].")
```
**Output**:
```json
[
  {'sequence': '[CLS] the woman worked as a teacher. [SEP]', 'token_str': 'teacher'},
  {'sequence': '[CLS] the woman worked as a nurse. [SEP]', 'token_str': 'nurse'},
  ...
]
```
These biases may propagate to downstream tasks. Due to its small size (~20 MB), the model is suitable for resource-constrained environments but may have limited capacity for complex tasks compared to larger variants.

### Recommendations

Users should:
- Conduct bias audits tailored to their application.
- Fine-tune with diverse, representative datasets to reduce bias.
- Apply model compression techniques (e.g., quantization) for deployment on ultra-constrained devices.

## How to Get Started with the Model

Use the code below to get started with the model.

```python
from transformers import pipeline, BertTokenizer, BertModel

# Masked Language Modeling
unmasker = pipeline('fill-mask', model='boltuix/bert-tinyplus')
result = unmasker("Hello I'm a [MASK] model.")
print(result)

# Feature Extraction (PyTorch)
tokenizer = BertTokenizer.from_pretrained('boltuix/bert-tinyplus')
model = BertModel.from_pretrained('boltuix/bert-tinyplus')
text = "Replace me by any text you'd like."
encoded_input = tokenizer(text, return_tensors='pt')
output = model(**encoded_input)
```

## Training Details

### Training Data

The model was pretrained on:
- **BookCorpus**: ~11,038 unpublished books, providing diverse narrative text.
- **English Wikipedia**: Excluding lists, tables, and headers for clean, factual content.

See the [BoltUIX Dataset Card](https://huggingface.co/boltuix/datasets) for more details.

### Training Procedure

#### Preprocessing

- Texts are lowercased and tokenized using WordPiece with a vocabulary size of 30,000.
- Inputs are formatted as: `[CLS] Sentence A [SEP] Sentence B [SEP]`.
- 50% of the time, Sentence A and B are consecutive; otherwise, Sentence B is random.
- Masking:
  - 15% of tokens are masked.
  - 80% of masked tokens are replaced with `[MASK]`.
  - 10% are replaced with a random token.
  - 10% are left unchanged.

#### Training Hyperparameters

- **Training regime:** fp16 mixed precision
- **Optimizer**: Adam (learning rate 1e-4, β1=0.9, β2=0.999, weight decay 0.01)
- **Batch size**: 64
- **Steps**: 500,000
- **Sequence length**: 128 tokens (98% of steps), 512 tokens (2% of steps)
- **Warmup**: 5,000 steps with linear learning rate decay

#### Speeds, Sizes, Times

- **Training time**: Approximately 60 hours
- **Checkpoint size**: ~20 MB
- **Throughput**: ~150 sentences/second on TPU infrastructure

## Evaluation

### Testing Data, Factors & Metrics

#### Testing Data

Evaluated on the GLUE benchmark, including tasks like MNLI, QQP, QNLI, SST-2, CoLA, STS-B, MRPC, and RTE.

#### Factors

- **Subpopulations**: General English text, academic, and professional domains
- **Domains**: News, books, Wikipedia, scientific articles

#### Metrics

- **Accuracy**: For classification tasks (e.g., MNLI, SST-2)
- **F1 Score**: For tasks like QQP, MRPC
- **Pearson/Spearman Correlation**: For STS-B

### Results

GLUE test results (fine-tuned):
| Task       | MNLI-(m/mm) | QQP  | QNLI | SST-2 | CoLA | STS-B | MRPC | RTE  | Average |
|------------|-------------|------|------|-------|------|-------|------|------|---------|
| Score      | 81.2/80.1   | 69.5 | 87.3 | 90.2  | 47.8 | 82.4  | 85.2 | 63.1 | 76.3    |

#### Summary

The model provides modest performance across GLUE tasks, with reasonable results in SST-2 and QNLI. It outperforms `boltuix/bert-micro` and `boltuix/bert-mini` in tasks like RTE and CoLA, offering slightly better capacity for lightweight applications.

## Model Examination

The model’s attention mechanisms were analyzed to ensure basic contextual understanding, with no significant overfitting observed during pretraining. Ablation studies validated the training configuration for lightweight, efficient performance.

## Environmental Impact

Carbon emissions estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) from [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).

- **Hardware Type**: 1 cloud TPU (4 TPU chips)
- **Hours used**: 60 hours
- **Cloud Provider**: Google Cloud
- **Compute Region**: us-central1
- **Carbon Emitted**: ~40 kg CO2eq (estimated based on TPU energy consumption and regional grid carbon intensity)

## Technical Specifications

### Model Architecture and Objective

- **Architecture**: BERT (transformer-based, bidirectional)
- **Objective**: Masked Language Modeling (MLM) and Next Sentence Prediction (NSP)
- **Layers**: 2
- **Hidden Size**: 256
- **Attention Heads**: 4

### Compute Infrastructure

#### Hardware

- 1 cloud TPU (4 TPU chips total)

#### Software

- PyTorch
- Transformers library (Hugging Face)

## Citation

**BibTeX:**
```bibtex
@article{DBLP:journals/corr/abs-1810-04805,
  author    = {Jacob Devlin and Ming{-}Wei Chang and Kenton Lee and Kristina Toutanova},
  title     = {{BERT:} Pre-training of Deep Bidirectional Transformers for Language Understanding},
  journal   = {CoRR},
  volume    = {abs/1810.04805},
  year      = {2018},
  url       = {http://arxiv.org/abs/1810.04805},
  archivePrefix = {arXiv},
  eprint    = {1810.04805}
}
```

**APA:**
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2018). BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. *CoRR, abs/1810.04805*. http://arxiv.org/abs/1810.04805

## Glossary

- **MLM**: Masked Language Modeling, where 15% of tokens are masked for prediction.
- **NSP**: Next Sentence Prediction, determining if two sentences are consecutive.
- **WordPiece**: Tokenization method splitting words into subword units.

## More Information

- See the [Hugging Face documentation](https://huggingface.co/docs/transformers/model_doc/bert) for advanced usage details.
- Contact: [email protected]

## Model Card Authors

- Hugging Face team
- BoltUIX contributors

## Model Card Contact

For questions, please contact [email protected] or open an issue on the [model repository](https://huggingface.co/boltuix/bert-tinyplus).