mathiascreutz
commited on
Commit
•
f275cf8
1
Parent(s):
48f737e
More comments
Browse files- opusparcus.py +12 -7
opusparcus.py
CHANGED
@@ -94,11 +94,12 @@ class Opusparcus(datasets.GeneratorBasedBuilder):
|
|
94 |
OpusparcusConfig(lang=lang, quality=quality, version=_VERSION) for lang in LANGS for quality in QUALITIES
|
95 |
]
|
96 |
|
97 |
-
#
|
|
|
98 |
|
99 |
def _info(self):
|
100 |
-
#
|
101 |
-
#
|
102 |
features = datasets.Features(
|
103 |
{
|
104 |
"lang": datasets.Value("string"),
|
@@ -113,11 +114,11 @@ class Opusparcus(datasets.GeneratorBasedBuilder):
|
|
113 |
# This is the description that will appear on the datasets page.
|
114 |
description=_DESCRIPTION,
|
115 |
# This defines the different columns of the dataset and their types
|
116 |
-
features=features,
|
117 |
# If there's a common (input, target) tuple from the features,
|
118 |
# specify them here. They'll be used if as_supervised=True in
|
119 |
-
# builder.as_dataset
|
120 |
-
supervised_keys=
|
121 |
# Homepage of the dataset for documentation
|
122 |
homepage=_HOMEPAGE,
|
123 |
# License for the dataset if available
|
@@ -137,6 +138,10 @@ class Opusparcus(datasets.GeneratorBasedBuilder):
|
|
137 |
# separated by a period, containing the values of
|
138 |
# self.config.lang and self.config.quality.
|
139 |
|
|
|
|
|
|
|
|
|
140 |
# Select which file of the training data contains the matching data:
|
141 |
if self.config.quality < 70:
|
142 |
# We need to retrieve the largest training set file
|
@@ -259,7 +264,7 @@ class Opusparcus(datasets.GeneratorBasedBuilder):
|
|
259 |
}
|
260 |
else:
|
261 |
# The validation and test sets are in jsonl files.
|
262 |
-
# They contain the fields "lang" and "
|
263 |
# If we ask for the full sets, we will keep all data entries, also
|
264 |
# the sentence pairs that were not considered paraphrases by the
|
265 |
# annotators:
|
|
|
94 |
OpusparcusConfig(lang=lang, quality=quality, version=_VERSION) for lang in LANGS for quality in QUALITIES
|
95 |
]
|
96 |
|
97 |
+
# There is no default configuration. User always needs to specify one:
|
98 |
+
#DEFAULT_CONFIG_NAME = None
|
99 |
|
100 |
def _info(self):
|
101 |
+
# This method specifies the datasets.DatasetInfo object which
|
102 |
+
# contains informations and typings for the dataset
|
103 |
features = datasets.Features(
|
104 |
{
|
105 |
"lang": datasets.Value("string"),
|
|
|
114 |
# This is the description that will appear on the datasets page.
|
115 |
description=_DESCRIPTION,
|
116 |
# This defines the different columns of the dataset and their types
|
117 |
+
features=features,
|
118 |
# If there's a common (input, target) tuple from the features,
|
119 |
# specify them here. They'll be used if as_supervised=True in
|
120 |
+
# builder.as_dataset:
|
121 |
+
supervised_keys=(sent1, sent2), # is this correct?
|
122 |
# Homepage of the dataset for documentation
|
123 |
homepage=_HOMEPAGE,
|
124 |
# License for the dataset if available
|
|
|
138 |
# separated by a period, containing the values of
|
139 |
# self.config.lang and self.config.quality.
|
140 |
|
141 |
+
if lang is None:
|
142 |
+
# This is an error, nothing to do here
|
143 |
+
return []
|
144 |
+
|
145 |
# Select which file of the training data contains the matching data:
|
146 |
if self.config.quality < 70:
|
147 |
# We need to retrieve the largest training set file
|
|
|
264 |
}
|
265 |
else:
|
266 |
# The validation and test sets are in jsonl files.
|
267 |
+
# They contain the fields "lang" and "annot_score" that we filter on.
|
268 |
# If we ask for the full sets, we will keep all data entries, also
|
269 |
# the sentence pairs that were not considered paraphrases by the
|
270 |
# annotators:
|