🐚 E3C Dataset for Transfer Learning, Weak Supervision and Cross-Lingual Explorations
for now, the dataset is located in this huggingface dataset repository: smeoni/e3c
🤫
To load the dataset in a training script, use this following:
from datasets import load_dataset
load_dataset("smeoni/e3c")
you can also load the dataset with a local path:
from datasets import load_dataset
load_dataset("./e3c/e3c.py")
The README.md inside the e3c folder is the metadata for the dataset. The entire e3c folder in this repository is uploaded in the huggingface repository. You can generate the metadata using the following command:
datasets-cli test e3c/e3c.py --save_infos --all_configs