|
import datasets |
|
|
|
features = datasets.Features( |
|
{ |
|
"source": datasets.Value(dtype="string", id=None), |
|
"id": datasets.Value(dtype="string", id=None), |
|
"text": datasets.Value(dtype="string", id=None), |
|
"added": datasets.Value(dtype="string", id=None), |
|
"timestamp": datasets.Value(dtype="timestamp[s]", id=None), |
|
"metadata": {"url": datasets.Value(dtype="string", id=None)}, |
|
} |
|
) |
|
dataset = datasets.load_dataset( |
|
"json", |
|
data_files={ |
|
"train": "fi_processed/c4-fi.*.json", |
|
"validation": "fi_processed/c4-fi-validation*.json", |
|
}, |
|
features=features, |
|
cache_dir="/researchdisk/datasets_cache", |
|
num_proc=96, |
|
) |
|
|
|
dataset = dataset.flatten() |
|
dataset = dataset.rename_column("metadata.url", "url") |
|
dataset = dataset.remove_columns(["source", "id", "added"]) |
|
print(dataset) |
|
dataset.save_to_disk("/researchdisk/mc4_3.1.0_fi", num_proc=96) |
|
|