Update README.md
Browse files
README.md
CHANGED
|
@@ -4,7 +4,15 @@
|
|
| 4 |
|
| 5 |
This repository contains all the reference datasets that are used for running evaluation on the GEM benchmark. These datasets were originally hosted as a [GitHub release](https://github.com/GEM-benchmark/GEM-metrics/releases) on the [`GEM-metrics`](https://github.com/GEM-benchmark/GEM-metrics) repository, but have been migrated to the Hugging Face Hub.
|
| 6 |
|
| 7 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 8 |
|
| 9 |
```python
|
| 10 |
python convert_datasets_to_json.py
|
|
|
|
| 4 |
|
| 5 |
This repository contains all the reference datasets that are used for running evaluation on the GEM benchmark. These datasets were originally hosted as a [GitHub release](https://github.com/GEM-benchmark/GEM-metrics/releases) on the [`GEM-metrics`](https://github.com/GEM-benchmark/GEM-metrics) repository, but have been migrated to the Hugging Face Hub.
|
| 6 |
|
| 7 |
+
## Converting datasets to JSON
|
| 8 |
+
|
| 9 |
+
We provide a `convert_dataset_to_json.py` conversion script that converts the datasets in the GEM organisation to the JSON format expected by the `GEM-metrics` library. To run the script, first install [`jq`](https://stedolan.github.io/jq/download/) and then install the script's Python dependencies:
|
| 10 |
+
|
| 11 |
+
```
|
| 12 |
+
python -m pip install -r requirements.txt
|
| 13 |
+
```
|
| 14 |
+
|
| 15 |
+
You can then run the script as follows:
|
| 16 |
|
| 17 |
```python
|
| 18 |
python convert_datasets_to_json.py
|