**Model Description:** The DistilBERT model was proposed in the blog post [Smaller, faster, cheaper, lighter: Introducing DistilBERT, adistilled version of BERT](https://medium.com/huggingface/distilbert-8cf3380435b5), and the paper [DistilBERT, adistilled version of BERT: smaller, faster, cheaper and lighter](https://arxiv.org/abs/1910.01108). DistilBERT is a small, fast, cheap and light Transformer model trained by distilling BERT base. It has 40% less parameters than *bert-base-uncased*, runs 60% faster while preserving over 95% of BERT's performances as measured on the GLUE language understanding benchmark.
This model is a fine-tune checkpoint of [DistilBERT-base-cased](https://huggingface.co/distilbert-base-cased), fine-tuned using (a second step of) knowledge distillation on [SQuAD v1.1](https://huggingface.co/datasets/squad).
- **Developed by:** Hugging Face
- **Model Type:** Transformer-based language model
- See [this repository](https://github.com/huggingface/transformers/tree/main/examples/research_projects/distillation) for more about Distil\* (a class of compressed models including this model)
- See [Sanh et al. (2019)](https://arxiv.org/abs/1910.01108) for more information about knowledge distillation and the training procedure
The model should not be used to intentionally create hostile or alienating environments for people. In addition, the model was not trained to be factual or true representations of people or events, and therefore using the model to generate such content is out-of-scope for the abilities of this model.
## Risks, Limitations and Biases
**CONTENT WARNING: Readers should be aware that language generated by this model can be disturbing or offensive to some and can propagate historical and current stereotypes.**
Significant research has explored bias and fairness issues with language models (see, e.g., [Sheng et al. (2021)](https://aclanthology.org/2021.acl-long.330.pdf) and [Bender et al. (2021)](https://dl.acm.org/doi/pdf/10.1145/3442188.3445922)). Predictions generated by the model can include disturbing and harmful stereotypes across protected classes; identity characteristics; and sensitive, social, and occupational groups. For example:
Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model.
## Training
#### Training Data
The [distilbert-base-cased model](https://huggingface.co/distilbert-base-cased) was trained using the same data as the [distilbert-base-uncased model](https://huggingface.co/distilbert-base-uncased). The [distilbert-base-uncased model](https://huggingface.co/distilbert-base-uncased) model describes it's training data as:
> DistilBERT pretrained on the same data as BERT, which is [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books and [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers).
To learn more about the SQuAD v1.1 dataset, see the [SQuAD v1.1 data card](https://huggingface.co/datasets/squad).
#### Training Procedure
##### Preprocessing
See the [distilbert-base-cased model card](https://huggingface.co/distilbert-base-cased) for further details.
##### Pretraining
See the [distilbert-base-cased model card](https://huggingface.co/distilbert-base-cased) for further details.
## Evaluation
As discussed in the [model repository](https://github.com/huggingface/transformers/blob/main/examples/research_projects/distillation/README.md)
> This model reaches a F1 score of 87.1 on the [SQuAD v1.1] dev set (for comparison, BERT bert-base-cased version reaches a F1 score of 88.7).
## Environmental Impact
Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700). We present the hardware type and hours used based on the [associated paper](https://arxiv.org/pdf/1910.01108.pdf). Note that these details are just for training DistilBERT, not including the fine-tuning with SQuAD.
- **Hardware Type:** 8 16GB V100 GPUs
- **Hours used:** 90 hours
- **Cloud Provider:** Unknown
- **Compute Region:** Unknown
- **Carbon Emitted:** Unknown
## Technical Specifications
See the [associated paper](https://arxiv.org/abs/1910.01108) for details on the modeling architecture, objective, compute infrastructure, and training details.
## Citation Information
```bibtex
@inproceedings{sanh2019distilbert,
title={DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter},
author={Sanh, Victor and Debut, Lysandre and Chaumond, Julien and Wolf, Thomas},
booktitle={NeurIPS EMC^2 Workshop},
year={2019}
}
```
APA:
- Sanh, V., Debut, L., Chaumond, J., & Wolf, T. (2019). DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter. arXiv preprint arXiv:1910.01108.
## Model Card Authors
This model card was written by the Hugging Face team.
