ScandEval 10.0.0

Evaluation of pretrained language models on mono- or multilingual language tasks.

Evaluation of pretrained language models on mono- or multilingual language tasks.

---

Maintainers

• Dan Saattrup Nielsen (@saattrupdan, [e-mail])
• Kenneth Enevoldsen (@KennethEnevoldsen, [e-mail])

Installation

To install the package simply write the following command in your favorite terminal:

$ pip install scandeval

Quickstart

Benchmarking from the Command Line

The easiest way to benchmark pretrained models is via the command line interface. After having installed the package, you can benchmark your favorite model like so:

$ scandeval --model-id <model-id>

Here model_id is the HuggingFace model ID, which can be found on the HuggingFace Hub. By default this will benchmark the model on all the datasets eligible. If you want to benchmark on a specific dataset, this can be done via the --dataset flag. This will for instance evaluate the model on the AngryTweets dataset:

$ scandeval --model-id <model-id> --dataset angry-tweets

We can also separate by language. To benchmark all Danish models on all Danish datasets, say, this can be done using the language tag, like so:

$ scandeval --language da

Multiple models, datasets and/or languages can be specified by just attaching multiple arguments. Here is an example with two models:

$ scandeval --model-id <model-id1> --model-id <model-id2> --dataset angry-tweets

The specific model version to use can also be added after the suffix '@':

$ scandeval --model-id <model-id>@<commit>

It can be a branch name, a tag name, or a commit id. It defaults to 'main' for latest.

See all the arguments and options available for the scandeval command by typing

$ scandeval --help

Benchmarking from a Script

In a script, the syntax is similar to the command line interface. You simply initialise an object of the Benchmarker class, and call this benchmark object with your favorite models and/or datasets:

>>> from scandeval import Benchmarker
>>> benchmark = Benchmarker()
>>> benchmark('<model-id>')

To benchmark on a specific dataset, you simply specify the second argument, shown here with the AngryTweets dataset again:

>>> benchmark('<model_id>', 'angry-tweets')

If you want to benchmark a subset of all the models on the Hugging Face Hub, you can specify several parameters in the Benchmarker initializer to narrow down the list of models to the ones you care about. As a simple example, the following would benchmark all the Nynorsk models on Nynorsk datasets:

>>> benchmark = Benchmarker(language='nn')
>>> benchmark()

Citing ScandEval

If you want to cite the framework then feel free to use this:

@inproceedings{nielsen2023scandeval,
  title={ScandEval: A Benchmark for Scandinavian Natural Language Processing},
  author={Nielsen, Dan Saattrup},
  booktitle={The 24rd Nordic Conference on Computational Linguistics},
  year={2023}
}

Remarks

The image used in the logo has been created by the amazing Scandinavia and the World team. Go check them out!