Release tagged datasets for LEACE paper on Hugging Face

Hi EleutherAI team 🤗

Niels here from the open-source team at Hugging Face. I discovered your work on LEACE through Hugging Face's daily papers: https://huggingface.co/papers/2306.03819. It's a very interesting approach to concept erasure!

The paper page lets people discuss your paper and find related artifacts. You can also claim the paper page to link it to your EleutherAI organization profile on HF.

I noticed in the `concept-erasure` repository README (`https://github.com/EleutherAI/concept-erasure`) that you plan to upload the "tagged datasets" used for the concept scrubbing experiments to the Hugging Face Hub shortly. That's great news, as hosting them on the Hub (https://huggingface.co/datasets) would make them easily accessible for the community!

This would allow users to load the dataset easily via our `datasets` library:

```python
from datasets import load_dataset

dataset = load_dataset("EleutherAI/your-tagged-dataset-name") # Or the final name
```

We have guides on uploading datasets here: https://huggingface.co/docs/datasets/loading. The [Dataset Viewer](https://huggingface.co/docs/hub/en/datasets-viewer) is also a nice feature for exploration.

Once uploaded, we can link the dataset repository to the paper page (https://huggingface.co/papers/2306.03819) for better discoverability.

Let us know if you need any help with the upload process or have any questions when the datasets are ready!

Cheers,

Niels
ML Engineer @ HF 🤗