Datasets:
Update mc4-sampling.py
Browse files- mc4-sampling.py +10 -12
mc4-sampling.py
CHANGED
@@ -271,17 +271,15 @@ _N_SHARDS_PER_SPLIT = {
|
|
271 |
class Mc4SamplingConfig(datasets.BuilderConfig):
|
272 |
"""BuilderConfig for mC4 Sampling."""
|
273 |
|
274 |
-
def __init__(self, *args, **kwargs):
|
275 |
"""BuilderConfig for mC4 Sampling.
|
276 |
Args:
|
277 |
languages (:obj:`List[str]`): list of languages to load
|
278 |
**kwargs: keyword arguments forwarded to super.
|
279 |
"""
|
280 |
-
print(args, kwargs)
|
281 |
-
languages = kwargs.pop("languages", [])
|
282 |
super().__init__(
|
283 |
*args,
|
284 |
-
|
285 |
**kwargs,
|
286 |
)
|
287 |
self.languages = languages
|
@@ -290,11 +288,11 @@ class Mc4SamplingConfig(datasets.BuilderConfig):
|
|
290 |
class Mc4Sampling(datasets.GeneratorBasedBuilder):
|
291 |
"""mC4 Sampling, a colossal, cleaned version of Common Crawl's web crawl corpus."""
|
292 |
|
293 |
-
BUILDER_CONFIGS = [Mc4SamplingConfig(
|
294 |
BUILDER_CONFIG_CLASS = Mc4SamplingConfig
|
295 |
|
296 |
-
def __init__(self, *args,
|
297 |
-
self.data_files = kwargs.
|
298 |
self.sampling_method = kwargs.pop("sampling_method", None)
|
299 |
self.perplexity_model = kwargs.pop("perplexity_model", None)
|
300 |
self.sampling_factor = kwargs.pop("sampling_factor", None)
|
@@ -322,11 +320,11 @@ class Mc4Sampling(datasets.GeneratorBasedBuilder):
|
|
322 |
self.should_keep_doc = self._should_keep_doc_gaussian
|
323 |
else:
|
324 |
self.should_keep_doc = self._should_keep_doc_step
|
325 |
-
init_kwargs = {
|
326 |
-
|
327 |
-
|
328 |
-
}
|
329 |
-
super().__init__(*args,
|
330 |
|
331 |
def get_perplexity(self, doc):
|
332 |
doc_log_score, doc_length = 0, 0
|
|
|
271 |
class Mc4SamplingConfig(datasets.BuilderConfig):
|
272 |
"""BuilderConfig for mC4 Sampling."""
|
273 |
|
274 |
+
def __init__(self, languages, *args, **kwargs):
|
275 |
"""BuilderConfig for mC4 Sampling.
|
276 |
Args:
|
277 |
languages (:obj:`List[str]`): list of languages to load
|
278 |
**kwargs: keyword arguments forwarded to super.
|
279 |
"""
|
|
|
|
|
280 |
super().__init__(
|
281 |
*args,
|
282 |
+
name="+".join(languages),
|
283 |
**kwargs,
|
284 |
)
|
285 |
self.languages = languages
|
|
|
288 |
class Mc4Sampling(datasets.GeneratorBasedBuilder):
|
289 |
"""mC4 Sampling, a colossal, cleaned version of Common Crawl's web crawl corpus."""
|
290 |
|
291 |
+
BUILDER_CONFIGS = [Mc4SamplingConfig(languages=[lang]) for lang in _LANGUAGES]
|
292 |
BUILDER_CONFIG_CLASS = Mc4SamplingConfig
|
293 |
|
294 |
+
def __init__(self, *args, **kwargs):
|
295 |
+
self.data_files = kwargs.get("data_files", {})
|
296 |
self.sampling_method = kwargs.pop("sampling_method", None)
|
297 |
self.perplexity_model = kwargs.pop("perplexity_model", None)
|
298 |
self.sampling_factor = kwargs.pop("sampling_factor", None)
|
|
|
320 |
self.should_keep_doc = self._should_keep_doc_gaussian
|
321 |
else:
|
322 |
self.should_keep_doc = self._should_keep_doc_step
|
323 |
+
# init_kwargs = {
|
324 |
+
# prop: kwargs.get(prop)
|
325 |
+
# for prop in ("name", "version", "data_dir", "data_files", "description")
|
326 |
+
# }
|
327 |
+
super().__init__(*args, **kwargs)
|
328 |
|
329 |
def get_perplexity(self, doc):
|
330 |
doc_log_score, doc_length = 0, 0
|