patrickvonplaten
commited on
Commit
•
52783d2
1
Parent(s):
bafe1b7
up
Browse files
de_head_0000_2015-49.txt.gz → dummy.txt.gz
RENAMED
File without changes
|
german_common_crawl.py
CHANGED
@@ -34,13 +34,14 @@ _CITATION = """\
|
|
34 |
_DESCRIPTION = """\
|
35 |
German Only Extract from Common Crawl
|
36 |
|
37 |
-
This Dataset is for pretraining a German Language Model (Unsupervised) or tune a Multilingual Model specifically to German
|
38 |
"""
|
39 |
|
40 |
REPO_URL = "https://huggingface.co/datasets/flax-community/german_common_crawl/resolve/main/"
|
41 |
|
42 |
_URL_FIRST = [REPO_URL + file_name for file_name in [
|
43 |
-
"de_head_0000_2015-48.txt.gz",
|
|
|
44 |
]]
|
45 |
|
46 |
#TODO convert & upload all those files correctly
|
@@ -153,12 +154,11 @@ _URL_HEAD = [REPO_URL + file_name for file_name in [
|
|
153 |
_URL_MIDDLE = [REPO_URL + file_name for file_name in [
|
154 |
]]
|
155 |
|
|
|
156 |
class GermanCommonCrawl(datasets.GeneratorBasedBuilder):
|
157 |
"""TODO: Short description of my dataset."""
|
158 |
|
159 |
VERSION = datasets.Version("1.1.0")
|
160 |
-
|
161 |
-
|
162 |
BUILDER_CONFIGS = [
|
163 |
datasets.BuilderConfig(name="first", version=VERSION, description="Only the first data file"),
|
164 |
datasets.BuilderConfig(name="head", version=VERSION, description=""), #TODO fill description
|
@@ -200,16 +200,16 @@ class GermanCommonCrawl(datasets.GeneratorBasedBuilder):
|
|
200 |
)
|
201 |
|
202 |
def _split_generators(self, dl_manager):
|
203 |
-
"""Returns SplitGenerators."""
|
204 |
-
if self.config.name == "first":
|
205 |
-
data_files = dl_manager.download(_URL_FIRST)
|
206 |
elif self.config.name == "head":
|
207 |
-
data_files = dl_manager.download(_URL_HEAD)
|
208 |
elif self.config.name == "middle":
|
209 |
-
data_files = dl_manager.download(_URL_MIDDLE)
|
210 |
else:
|
211 |
-
data_files = dl_manager.download(_URL_HEAD + _URL_MIDDLE)
|
212 |
-
|
213 |
return [
|
214 |
datasets.SplitGenerator(
|
215 |
name=datasets.Split.TRAIN,
|
@@ -221,9 +221,8 @@ class GermanCommonCrawl(datasets.GeneratorBasedBuilder):
|
|
221 |
|
222 |
def _generate_examples(self, data_files):
|
223 |
"""This function returns the examples in the raw (text) form by iterating on all the files."""
|
224 |
-
for filepath in data_files:
|
225 |
with gzip.open(filepath, "rt", encoding="utf-8") as f:
|
226 |
-
import ipdb; ipdb.set_trace()
|
227 |
for id_, line in enumerate(f):
|
228 |
item = literal_eval(line)
|
229 |
yield id_, {
|
@@ -239,4 +238,7 @@ class GermanCommonCrawl(datasets.GeneratorBasedBuilder):
|
|
239 |
"original_nlines": item["original_nlines"],
|
240 |
"original_length": item["original_length"],
|
241 |
"language": item["language"],
|
|
|
|
|
|
|
242 |
}
|
|
|
34 |
_DESCRIPTION = """\
|
35 |
German Only Extract from Common Crawl
|
36 |
|
37 |
+
This Dataset is for pretraining a German Language Model (Unsupervised) or tune a Multilingual Model specifically to German
|
38 |
"""
|
39 |
|
40 |
REPO_URL = "https://huggingface.co/datasets/flax-community/german_common_crawl/resolve/main/"
|
41 |
|
42 |
_URL_FIRST = [REPO_URL + file_name for file_name in [
|
43 |
+
# "de_head_0000_2015-48.txt.gz",
|
44 |
+
"dummy.txt.gz",
|
45 |
]]
|
46 |
|
47 |
#TODO convert & upload all those files correctly
|
|
|
154 |
_URL_MIDDLE = [REPO_URL + file_name for file_name in [
|
155 |
]]
|
156 |
|
157 |
+
|
158 |
class GermanCommonCrawl(datasets.GeneratorBasedBuilder):
|
159 |
"""TODO: Short description of my dataset."""
|
160 |
|
161 |
VERSION = datasets.Version("1.1.0")
|
|
|
|
|
162 |
BUILDER_CONFIGS = [
|
163 |
datasets.BuilderConfig(name="first", version=VERSION, description="Only the first data file"),
|
164 |
datasets.BuilderConfig(name="head", version=VERSION, description=""), #TODO fill description
|
|
|
200 |
)
|
201 |
|
202 |
def _split_generators(self, dl_manager):
|
203 |
+
"""Returns SplitGenerators."""
|
204 |
+
if self.config.name == "first":
|
205 |
+
data_files = dl_manager.download(_URL_FIRST)
|
206 |
elif self.config.name == "head":
|
207 |
+
data_files = dl_manager.download(_URL_HEAD)
|
208 |
elif self.config.name == "middle":
|
209 |
+
data_files = dl_manager.download(_URL_MIDDLE)
|
210 |
else:
|
211 |
+
data_files = dl_manager.download(_URL_HEAD + _URL_MIDDLE)
|
212 |
+
|
213 |
return [
|
214 |
datasets.SplitGenerator(
|
215 |
name=datasets.Split.TRAIN,
|
|
|
221 |
|
222 |
def _generate_examples(self, data_files):
|
223 |
"""This function returns the examples in the raw (text) form by iterating on all the files."""
|
224 |
+
for filepath in data_files:
|
225 |
with gzip.open(filepath, "rt", encoding="utf-8") as f:
|
|
|
226 |
for id_, line in enumerate(f):
|
227 |
item = literal_eval(line)
|
228 |
yield id_, {
|
|
|
238 |
"original_nlines": item["original_nlines"],
|
239 |
"original_length": item["original_length"],
|
240 |
"language": item["language"],
|
241 |
+
"language_score": item["language_score"],
|
242 |
+
"perplexity": item["perplexity"],
|
243 |
+
"bucket": item["bucket"],
|
244 |
}
|