Commit
·
6d98a07
1
Parent(s):
791c5c3
Support streaming xtreme dataset for PAN-X config (#4135)
Browse filesCommit from https://github.com/huggingface/datasets/commit/9977ade72191ff0b6907ec63935448c6269a91a1
xtreme.py
CHANGED
|
@@ -471,32 +471,12 @@ class Xtreme(datasets.GeneratorBasedBuilder):
|
|
| 471 |
)
|
| 472 |
if self.config.name.startswith("PAWS-X"):
|
| 473 |
features = PawsxParser.features
|
| 474 |
-
|
| 475 |
features["gold_label"] = datasets.Value("string")
|
| 476 |
-
|
| 477 |
-
if self.config.name.startswith("udpos"):
|
| 478 |
features = UdposParser.features
|
| 479 |
-
|
| 480 |
-
|
| 481 |
-
features = datasets.Features(
|
| 482 |
-
{
|
| 483 |
-
"tokens": datasets.Sequence(datasets.Value("string")),
|
| 484 |
-
"ner_tags": datasets.Sequence(
|
| 485 |
-
datasets.features.ClassLabel(
|
| 486 |
-
names=[
|
| 487 |
-
"O",
|
| 488 |
-
"B-PER",
|
| 489 |
-
"I-PER",
|
| 490 |
-
"B-ORG",
|
| 491 |
-
"I-ORG",
|
| 492 |
-
"B-LOC",
|
| 493 |
-
"I-LOC",
|
| 494 |
-
]
|
| 495 |
-
)
|
| 496 |
-
),
|
| 497 |
-
"langs": datasets.Sequence(datasets.Value("string")),
|
| 498 |
-
}
|
| 499 |
-
)
|
| 500 |
return datasets.DatasetInfo(
|
| 501 |
# This is the description that will appear on the datasets page.
|
| 502 |
description=self.config.description + "\n" + _DESCRIPTION,
|
|
@@ -642,27 +622,7 @@ class Xtreme(datasets.GeneratorBasedBuilder):
|
|
| 642 |
]
|
| 643 |
|
| 644 |
if self.config.name.startswith("PAN-X"):
|
| 645 |
-
|
| 646 |
-
lang = self.config.name.split(".")[1]
|
| 647 |
-
lang_folder = dl_manager.extract(os.path.join(panx_dl_dir, lang + ".tar.gz"))
|
| 648 |
-
|
| 649 |
-
return [
|
| 650 |
-
datasets.SplitGenerator(
|
| 651 |
-
name=datasets.Split.VALIDATION,
|
| 652 |
-
# These kwargs will be passed to _generate_examples
|
| 653 |
-
gen_kwargs={"filepath": os.path.join(lang_folder, "dev")},
|
| 654 |
-
),
|
| 655 |
-
datasets.SplitGenerator(
|
| 656 |
-
name=datasets.Split.TEST,
|
| 657 |
-
# These kwargs will be passed to _generate_examples
|
| 658 |
-
gen_kwargs={"filepath": os.path.join(lang_folder, "test")},
|
| 659 |
-
),
|
| 660 |
-
datasets.SplitGenerator(
|
| 661 |
-
name=datasets.Split.TRAIN,
|
| 662 |
-
# These kwargs will be passed to _generate_examples
|
| 663 |
-
gen_kwargs={"filepath": os.path.join(lang_folder, "train")},
|
| 664 |
-
),
|
| 665 |
-
]
|
| 666 |
|
| 667 |
def _generate_examples(self, filepath=None, **kwargs):
|
| 668 |
"""Yields examples."""
|
|
@@ -784,20 +744,70 @@ class Xtreme(datasets.GeneratorBasedBuilder):
|
|
| 784 |
if self.config.name.startswith("udpos"):
|
| 785 |
yield from UdposParser.generate_examples(config=self.config, filepath=filepath, **kwargs)
|
| 786 |
if self.config.name.startswith("PAN-X"):
|
| 787 |
-
|
| 788 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 789 |
tokens = []
|
| 790 |
ner_tags = []
|
| 791 |
langs = []
|
| 792 |
-
for line in
|
|
|
|
| 793 |
if line == "" or line == "\n":
|
| 794 |
if tokens:
|
| 795 |
-
yield
|
| 796 |
"tokens": tokens,
|
| 797 |
"ner_tags": ner_tags,
|
| 798 |
"langs": langs,
|
| 799 |
}
|
| 800 |
-
|
| 801 |
tokens = []
|
| 802 |
ner_tags = []
|
| 803 |
langs = []
|
|
@@ -813,7 +823,7 @@ class Xtreme(datasets.GeneratorBasedBuilder):
|
|
| 813 |
# examples have no label in test set
|
| 814 |
ner_tags.append("O")
|
| 815 |
if tokens:
|
| 816 |
-
yield
|
| 817 |
"tokens": tokens,
|
| 818 |
"ner_tags": ner_tags,
|
| 819 |
"langs": langs,
|
|
|
|
| 471 |
)
|
| 472 |
if self.config.name.startswith("PAWS-X"):
|
| 473 |
features = PawsxParser.features
|
| 474 |
+
elif self.config.name == "XNLI":
|
| 475 |
features["gold_label"] = datasets.Value("string")
|
| 476 |
+
elif self.config.name.startswith("udpos"):
|
|
|
|
| 477 |
features = UdposParser.features
|
| 478 |
+
elif self.config.name.startswith("PAN-X"):
|
| 479 |
+
features = PanxParser.features
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 480 |
return datasets.DatasetInfo(
|
| 481 |
# This is the description that will appear on the datasets page.
|
| 482 |
description=self.config.description + "\n" + _DESCRIPTION,
|
|
|
|
| 622 |
]
|
| 623 |
|
| 624 |
if self.config.name.startswith("PAN-X"):
|
| 625 |
+
return PanxParser.split_generators(dl_manager=dl_manager, config=self.config)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 626 |
|
| 627 |
def _generate_examples(self, filepath=None, **kwargs):
|
| 628 |
"""Yields examples."""
|
|
|
|
| 744 |
if self.config.name.startswith("udpos"):
|
| 745 |
yield from UdposParser.generate_examples(config=self.config, filepath=filepath, **kwargs)
|
| 746 |
if self.config.name.startswith("PAN-X"):
|
| 747 |
+
yield from PanxParser.generate_examples(filepath=filepath, **kwargs)
|
| 748 |
+
|
| 749 |
+
|
| 750 |
+
class PanxParser:
|
| 751 |
+
|
| 752 |
+
features = datasets.Features(
|
| 753 |
+
{
|
| 754 |
+
"tokens": datasets.Sequence(datasets.Value("string")),
|
| 755 |
+
"ner_tags": datasets.Sequence(
|
| 756 |
+
datasets.features.ClassLabel(
|
| 757 |
+
names=[
|
| 758 |
+
"O",
|
| 759 |
+
"B-PER",
|
| 760 |
+
"I-PER",
|
| 761 |
+
"B-ORG",
|
| 762 |
+
"I-ORG",
|
| 763 |
+
"B-LOC",
|
| 764 |
+
"I-LOC",
|
| 765 |
+
]
|
| 766 |
+
)
|
| 767 |
+
),
|
| 768 |
+
"langs": datasets.Sequence(datasets.Value("string")),
|
| 769 |
+
}
|
| 770 |
+
)
|
| 771 |
+
|
| 772 |
+
@staticmethod
|
| 773 |
+
def split_generators(dl_manager=None, config=None):
|
| 774 |
+
data_dir = dl_manager.download_and_extract(config.data_url)
|
| 775 |
+
lang = config.name.split(".")[1]
|
| 776 |
+
archive = os.path.join(data_dir, lang + ".tar.gz")
|
| 777 |
+
split_filenames = {
|
| 778 |
+
datasets.Split.TRAIN: "train",
|
| 779 |
+
datasets.Split.VALIDATION: "dev",
|
| 780 |
+
datasets.Split.TEST: "test",
|
| 781 |
+
}
|
| 782 |
+
return [
|
| 783 |
+
datasets.SplitGenerator(
|
| 784 |
+
name=split,
|
| 785 |
+
gen_kwargs={
|
| 786 |
+
"filepath": dl_manager.iter_archive(archive),
|
| 787 |
+
"filename": split_filenames[split],
|
| 788 |
+
},
|
| 789 |
+
)
|
| 790 |
+
for split in split_filenames
|
| 791 |
+
]
|
| 792 |
+
|
| 793 |
+
@staticmethod
|
| 794 |
+
def generate_examples(filepath=None, filename=None):
|
| 795 |
+
idx = 1
|
| 796 |
+
for path, file in filepath:
|
| 797 |
+
if path.endswith(filename):
|
| 798 |
tokens = []
|
| 799 |
ner_tags = []
|
| 800 |
langs = []
|
| 801 |
+
for line in file:
|
| 802 |
+
line = line.decode("utf-8")
|
| 803 |
if line == "" or line == "\n":
|
| 804 |
if tokens:
|
| 805 |
+
yield idx, {
|
| 806 |
"tokens": tokens,
|
| 807 |
"ner_tags": ner_tags,
|
| 808 |
"langs": langs,
|
| 809 |
}
|
| 810 |
+
idx += 1
|
| 811 |
tokens = []
|
| 812 |
ner_tags = []
|
| 813 |
langs = []
|
|
|
|
| 823 |
# examples have no label in test set
|
| 824 |
ner_tags.append("O")
|
| 825 |
if tokens:
|
| 826 |
+
yield idx, {
|
| 827 |
"tokens": tokens,
|
| 828 |
"ner_tags": ner_tags,
|
| 829 |
"langs": langs,
|