Upload create_dataset.py
Browse files- create_dataset.py +34 -0
create_dataset.py
ADDED
@@ -0,0 +1,34 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from datasets import load_dataset
|
2 |
+
from huggingface_hub import create_repo, Repository, upload_file
|
3 |
+
import os
|
4 |
+
import typer
|
5 |
+
|
6 |
+
def main(language_label):
|
7 |
+
raw_data = load_dataset("AmazonScience/massive", language_label)
|
8 |
+
raw_data = raw_data.rename_column("utt", "text")
|
9 |
+
raw_data = raw_data.rename_column("intent", "label")
|
10 |
+
raw_data = raw_data.remove_columns(["locale", "partition", "scenario", "annot_utt",
|
11 |
+
"slot_method", "worker_id", "judgments"])
|
12 |
+
|
13 |
+
#to get labels
|
14 |
+
labels = raw_data["train"].features["label"]
|
15 |
+
|
16 |
+
#for uploading to hub
|
17 |
+
repo_name = "amazon_massive_intent_" + language_label
|
18 |
+
create_repo(repo_name, organization="SetFit", repo_type="dataset")
|
19 |
+
|
20 |
+
for split, dataset in raw_data.items():
|
21 |
+
dataset = dataset.map(lambda x: {"label_text": labels.int2str(x["label"])}, num_proc=4)
|
22 |
+
dataset.to_json(f"{split}.jsonl")
|
23 |
+
upload_file(f"{split}.jsonl", path_in_repo=f"{split}.jsonl", repo_id="SetFit/" + repo_name, repo_type="dataset")
|
24 |
+
os.system(f"rm {split}.jsonl")
|
25 |
+
|
26 |
+
upload_file("create_dataset.py", path_in_repo="create_dataset.py", repo_id="SetFit/" + repo_name, repo_type="dataset")
|
27 |
+
|
28 |
+
|
29 |
+
|
30 |
+
|
31 |
+
if __name__ == "__main__":
|
32 |
+
typer.run(main)
|
33 |
+
|
34 |
+
|