riaz.somc/transformers
1
0
Fork 0
mirror of https://github.com/huggingface/transformers.git synced 2025-08-01 18:51:14 +06:00
Code Issues Actions 3 Packages Projects Releases Wiki Activity

[s2s] wmt download script use less ram (#6405)

Browse Source
This commit is contained in:
Stas Bekman 2020-08-11 09:04:17 -07:00 committed by GitHub
parent 7c6a085ebf
commit f6cb0f806e
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 21 additions and 17 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

38
examples/seq2seq/download_wmt.py
View File

@ -4,44 +4,48 @@ import fire
from tqdm import tqdm from tqdm import tqdm
def download_wmt_dataset(src_lang, tgt_lang, dataset="wmt19", save_dir=None) -> None: def download_wmt_dataset(src_lang="ro", tgt_lang="en", dataset="wmt16", save_dir=None) -> None:
"""Download a dataset using the nlp package and save it to the format expected by finetune.py """Download a dataset using the nlp package and save it to the format expected by finetune.py
Format of save_dir: train.source, train.target, val.source, val.target, test.source, test.target. Format of save_dir: train.source, train.target, val.source, val.target, test.source, test.target.
Args: Args:
src_lang: <str> source language src_lang: <str> source language
tgt_lang: <str> target language tgt_lang: <str> target language
dataset: <str> like wmt19 (if you don't know, try wmt19). dataset: <str> wmt16, wmt17, etc. wmt16 is a good start as it's small. To get the full list run `import nlp; print([d.id for d in nlp.list_datasets() if "wmt" in d.id])`
save_dir: <str>, where to save the datasets, defaults to f'{dataset}-{src_lang}-{tgt_lang}' save_dir: <str>, where to save the datasets, defaults to f'{dataset}-{src_lang}-{tgt_lang}'
Usage: Usage:
>>> download_wmt_dataset('en', 'ru', dataset='wmt19') # saves to wmt19_en_ru >>> download_wmt_dataset('ro', 'en', dataset='wmt16') # saves to wmt16-ro-en
""" """
try: try:
import nlp import nlp
except (ModuleNotFoundError, ImportError): except (ModuleNotFoundError, ImportError):
raise ImportError("run pip install nlp") raise ImportError("run pip install nlp")
pair = f"{src_lang}-{tgt_lang}" pair = f"{src_lang}-{tgt_lang}"
print(f"Converting {dataset}-{pair}")
ds = nlp.load_dataset(dataset, pair) ds = nlp.load_dataset(dataset, pair)
if save_dir is None: if save_dir is None:
save_dir = f"{dataset}-{pair}" save_dir = f"{dataset}-{pair}"
save_dir = Path(save_dir) save_dir = Path(save_dir)
save_dir.mkdir(exist_ok=True) save_dir.mkdir(exist_ok=True)
for split in tqdm(ds.keys()): for split in ds.keys():
tr_list = list(ds[split]) print(f"Splitting {split} with {ds[split].num_rows} records")
data = [x["translation"] for x in tr_list]
src, tgt = [], [] # to save to val.source, val.target like summary datasets
for example in data: fn = "val" if split == "validation" else split
src.append(example[src_lang]) src_path = save_dir.joinpath(f"{fn}.source")
tgt.append(example[tgt_lang]) tgt_path = save_dir.joinpath(f"{fn}.target")
if split == "validation": src_fp = src_path.open("w+")
split = "val" # to save to val.source, val.target like summary datasets tgt_fp = tgt_path.open("w+")
src_path = save_dir.joinpath(f"{split}.source")
src_path.open("w+").write("\n".join(src)) # reader is the bottleneck so writing one record at a time doesn't slow things down
tgt_path = save_dir.joinpath(f"{split}.target") for x in tqdm(ds[split]):
tgt_path.open("w+").write("\n".join(tgt)) ex = x["translation"]
print(f"saved dataset to {save_dir}") src_fp.write(ex[src_lang] + "\n")
tgt_fp.write(ex[tgt_lang] + "\n")
print(f"Saved {dataset} dataset to {save_dir}")
if __name__ == "__main__": if __name__ == "__main__":