---
language: en
license: apache-2.0
---

# Model Card for distilroberta-base-climate-d-s

## Model Description

This is the ClimateBERT language model based on the DIV-SELECT and SIM-SELECT sample selection strategy.

*Note: We generally recommend choosing the [distilroberta-base-climate-f](https://huggingface.co/climatebert/distilroberta-base-climate-f) language model over this language model (unless you have good reasons not to).*

Using the [DistilRoBERTa](https://huggingface.co/distilroberta-base) model as starting point, the ClimateBERT Language Model is additionally pre-trained on a text corpus comprising climate-related research paper abstracts, corporate and general news and reports from companies. The underlying methodology can be found in our [language model research paper](https://arxiv.org/abs/2110.12010).

## Climate performance card

| distilroberta-base-climate-d-s                                             |                |
|--------------------------------------------------------------------------|----------------|
| 1. Is the resulting model publicly available?                            | Yes            |
| 2. How much time does the training of the final model take?              | 48 hours        |
| 3. How much time did all experiments take (incl. hyperparameter search)? | 350 hours      |
| 4. What was the power of GPU and CPU?                                    | 0.7 kW         |
| 5. At which geo location were the computations performed?                | Germany        |
| 6. What was the energy mix at the geo location?                          | 470 gCO2eq/kWh |
| 7. How much CO2eq was emitted to train the final model?                  | 15.79 kg        |
| 8. How much CO2eq was emitted for all experiments?                       | 115.15 kg       |
| 9. What is the average CO2eq emission for the inference of one sample?   | 0.62 mg        |
| 10. Which positive environmental impact can be expected from this work?  | This work can be categorized as a building block tools following Jin et al (2021). It supports the training of NLP models in the field of climate change and, thereby, have a positive environmental impact in the future. |
| 11. Comments                                                             | Block pruning could decrease CO2eq emissions |

## Citation Information

```bibtex
@article{wkbl2021,
        title={ClimateBERT: A Pretrained Language Model for Climate-Related Text},
        author={Webersinke, Nicolas and Kraus, Mathias and Bingler, Julia and Leippold, Markus},
        journal={arXiv preprint arXiv:2110.12010},
        year={2021}
}
```