VictorSanh
commited on
Commit
•
856a31b
1
Parent(s):
210a627
breaking down download of files
Browse files
P3.py
CHANGED
@@ -41,11 +41,50 @@ _HOMEPAGE = "https://github.com/bigscience-workshop/promptsource"
|
|
41 |
_DATA_PATH = "data"
|
42 |
|
43 |
|
44 |
-
def load_cached_task(cache_dir, split):
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
49 |
|
50 |
# Use `FixedLenSequenceFeature` for sequences with variable length.
|
51 |
def _feature_config(shape, dtype):
|
@@ -62,10 +101,10 @@ def load_cached_task(cache_dir, split):
|
|
62 |
feat: _feature_config(**desc) for feat, desc in features.items()
|
63 |
}
|
64 |
|
65 |
-
tfrecords = os.path.join(
|
66 |
-
|
67 |
-
)
|
68 |
-
ds = tf.data.TFRecordDataset(tf.io.gfile.glob(
|
69 |
ds = ds.map(
|
70 |
lambda pb: tf.io.parse_single_example(pb, feature_description),
|
71 |
num_parallel_calls=tf.data.experimental.AUTOTUNE
|
@@ -78,7 +117,6 @@ def load_cached_task(cache_dir, split):
|
|
78 |
)
|
79 |
return ds
|
80 |
|
81 |
-
|
82 |
def find_task_splits_and_features():
|
83 |
"""Find the available tasks under ./data and their available splits and features."""
|
84 |
task_and_their_splits = defaultdict(dict)
|
@@ -100,6 +138,7 @@ def find_task_splits_and_features():
|
|
100 |
with open(os.path.join(folder_path, f"info.{split_name}.json")) as f:
|
101 |
split_info = json.load(f)
|
102 |
features = split_info["features"]
|
|
|
103 |
|
104 |
# All splits under the same task have the same features dictionary (and thus the same features list)
|
105 |
if task_and_their_splits[task_name] == {}:
|
@@ -118,7 +157,16 @@ def find_task_splits_and_features():
|
|
118 |
|
119 |
|
120 |
_TASK_SPLITS_AND_FEATURES = find_task_splits_and_features()
|
121 |
-
_URLs = {
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
122 |
|
123 |
|
124 |
class P3Config(datasets.BuilderConfig):
|
@@ -184,13 +232,13 @@ class P3(datasets.GeneratorBasedBuilder):
|
|
184 |
def _split_generators(self, dl_manager):
|
185 |
split_generators = []
|
186 |
data_dir = dl_manager.download_and_extract(_URLs)
|
187 |
-
import pdb; pdb.set_trace()
|
188 |
if "train" in self.config.splits:
|
189 |
split_generators.append(
|
190 |
datasets.SplitGenerator(
|
191 |
name=datasets.Split.TRAIN,
|
192 |
gen_kwargs={
|
193 |
-
"
|
|
|
194 |
"split": "train",
|
195 |
}
|
196 |
)
|
@@ -200,7 +248,8 @@ class P3(datasets.GeneratorBasedBuilder):
|
|
200 |
datasets.SplitGenerator(
|
201 |
name=datasets.Split.VALIDATION,
|
202 |
gen_kwargs={
|
203 |
-
"
|
|
|
204 |
"split": "validation",
|
205 |
}
|
206 |
)
|
@@ -210,7 +259,8 @@ class P3(datasets.GeneratorBasedBuilder):
|
|
210 |
datasets.SplitGenerator(
|
211 |
name=datasets.Split.TEST,
|
212 |
gen_kwargs={
|
213 |
-
"
|
|
|
214 |
"split": "test",
|
215 |
}
|
216 |
)
|
@@ -222,7 +272,8 @@ class P3(datasets.GeneratorBasedBuilder):
|
|
222 |
datasets.SplitGenerator(
|
223 |
name=datasets.Split(special_split_name),
|
224 |
gen_kwargs={
|
225 |
-
"
|
|
|
226 |
"split": special_split_name,
|
227 |
}
|
228 |
)
|
@@ -230,7 +281,7 @@ class P3(datasets.GeneratorBasedBuilder):
|
|
230 |
return split_generators
|
231 |
|
232 |
|
233 |
-
def _generate_examples(self,
|
234 |
"""This function returns the examples in the raw (text) form."""
|
235 |
_FEAT_MAPPING_FUNCTIONS = {
|
236 |
"answer_choices": lambda x: [choice.decode("utf-8") for choice in x],
|
@@ -244,7 +295,7 @@ class P3(datasets.GeneratorBasedBuilder):
|
|
244 |
}
|
245 |
|
246 |
key = 0
|
247 |
-
ds = load_cached_task(
|
248 |
for ex in ds.as_numpy_iterator():
|
249 |
ex_dict = {}
|
250 |
for feat_name, feat_value in ex.items():
|
|
|
41 |
_DATA_PATH = "data"
|
42 |
|
43 |
|
44 |
+
# def load_cached_task(cache_dir, split):
|
45 |
+
# # TODO(Victor): this info.*.json is actually done twice... -> factorize
|
46 |
+
# with tf.io.gfile.GFile(os.path.join(cache_dir, f"info.{split}.json")) as f:
|
47 |
+
# split_info = json.load(f)
|
48 |
+
# features = split_info["features"]
|
49 |
+
|
50 |
+
# # Use `FixedLenSequenceFeature` for sequences with variable length.
|
51 |
+
# def _feature_config(shape, dtype):
|
52 |
+
# if dtype in ("int32", "bool"):
|
53 |
+
# # int32 and bool are stored as int64 in the tf.train.Example protobuf.
|
54 |
+
# dtype = "int64"
|
55 |
+
# if shape and shape[0] is None:
|
56 |
+
# return tf.io.FixedLenSequenceFeature(
|
57 |
+
# shape[1:], dtype, allow_missing=True
|
58 |
+
# )
|
59 |
+
# return tf.io.FixedLenFeature(shape, dtype)
|
60 |
+
|
61 |
+
# feature_description = {
|
62 |
+
# feat: _feature_config(**desc) for feat, desc in features.items()
|
63 |
+
# }
|
64 |
+
|
65 |
+
# tfrecords = os.path.join(
|
66 |
+
# cache_dir, f"{split}.tfrecord-*-of-*{split_info['num_shards']}"
|
67 |
+
# )
|
68 |
+
# ds = tf.data.TFRecordDataset(tf.io.gfile.glob(tfrecords))
|
69 |
+
# ds = ds.map(
|
70 |
+
# lambda pb: tf.io.parse_single_example(pb, feature_description),
|
71 |
+
# num_parallel_calls=tf.data.experimental.AUTOTUNE
|
72 |
+
# )
|
73 |
+
# # Cast features back to the types from the info JSON since some features
|
74 |
+
# # must be cast for storage (e.g., in32 is stored as int64).
|
75 |
+
# ds = ds.map(
|
76 |
+
# lambda x: {k: tf.cast(v, features[k]["dtype"]) for k, v in x.items()},
|
77 |
+
# num_parallel_calls=tf.data.experimental.AUTOTUNE
|
78 |
+
# )
|
79 |
+
# return ds
|
80 |
+
|
81 |
+
def load_cached_task(features_file, tfrecord, split):
|
82 |
+
# # TODO(Victor): this info.*.json is actually done twice... -> factorize
|
83 |
+
# with tf.io.gfile.GFile(os.path.join(cache_dir, f"info.{split}.json")) as f:
|
84 |
+
# split_info = json.load(f)
|
85 |
+
# features = split_info["features"]
|
86 |
+
with tf.io.gfile.GFile(features_file) as f:
|
87 |
+
features = json.load(f)
|
88 |
|
89 |
# Use `FixedLenSequenceFeature` for sequences with variable length.
|
90 |
def _feature_config(shape, dtype):
|
|
|
101 |
feat: _feature_config(**desc) for feat, desc in features.items()
|
102 |
}
|
103 |
|
104 |
+
# tfrecords = os.path.join(
|
105 |
+
# cache_dir, f"{split}.tfrecord-*-of-*{split_info['num_shards']}"
|
106 |
+
# )
|
107 |
+
ds = tf.data.TFRecordDataset(tf.io.gfile.glob([tfrecord]))
|
108 |
ds = ds.map(
|
109 |
lambda pb: tf.io.parse_single_example(pb, feature_description),
|
110 |
num_parallel_calls=tf.data.experimental.AUTOTUNE
|
|
|
117 |
)
|
118 |
return ds
|
119 |
|
|
|
120 |
def find_task_splits_and_features():
|
121 |
"""Find the available tasks under ./data and their available splits and features."""
|
122 |
task_and_their_splits = defaultdict(dict)
|
|
|
138 |
with open(os.path.join(folder_path, f"info.{split_name}.json")) as f:
|
139 |
split_info = json.load(f)
|
140 |
features = split_info["features"]
|
141 |
+
assert split_info["num_shards"] == 1
|
142 |
|
143 |
# All splits under the same task have the same features dictionary (and thus the same features list)
|
144 |
if task_and_their_splits[task_name] == {}:
|
|
|
157 |
|
158 |
|
159 |
_TASK_SPLITS_AND_FEATURES = find_task_splits_and_features()
|
160 |
+
_URLs = {
|
161 |
+
task_name: {
|
162 |
+
split_name: {
|
163 |
+
"tfrecord": f"{_DATA_PATH}/{task_name}/{split_name}.tfrecord-00000-of-00001",
|
164 |
+
"features_file": f"{_DATA_PATH}/{task_name}/info.{split_name}.json",
|
165 |
+
}
|
166 |
+
for split_name in splits_and_features["splits"]
|
167 |
+
}
|
168 |
+
for task_name, splits_and_features in _TASK_SPLITS_AND_FEATURES.items()
|
169 |
+
}
|
170 |
|
171 |
|
172 |
class P3Config(datasets.BuilderConfig):
|
|
|
232 |
def _split_generators(self, dl_manager):
|
233 |
split_generators = []
|
234 |
data_dir = dl_manager.download_and_extract(_URLs)
|
|
|
235 |
if "train" in self.config.splits:
|
236 |
split_generators.append(
|
237 |
datasets.SplitGenerator(
|
238 |
name=datasets.Split.TRAIN,
|
239 |
gen_kwargs={
|
240 |
+
"features_file": data_dir["features_file"],
|
241 |
+
"tfrecord": data_dir["tfrecord"],
|
242 |
"split": "train",
|
243 |
}
|
244 |
)
|
|
|
248 |
datasets.SplitGenerator(
|
249 |
name=datasets.Split.VALIDATION,
|
250 |
gen_kwargs={
|
251 |
+
"features_file": data_dir["features_file"],
|
252 |
+
"tfrecord": data_dir["tfrecord"],
|
253 |
"split": "validation",
|
254 |
}
|
255 |
)
|
|
|
259 |
datasets.SplitGenerator(
|
260 |
name=datasets.Split.TEST,
|
261 |
gen_kwargs={
|
262 |
+
"features_file": data_dir["features_file"],
|
263 |
+
"tfrecord": data_dir["tfrecord"],
|
264 |
"split": "test",
|
265 |
}
|
266 |
)
|
|
|
272 |
datasets.SplitGenerator(
|
273 |
name=datasets.Split(special_split_name),
|
274 |
gen_kwargs={
|
275 |
+
"features_file": data_dir["features_file"],
|
276 |
+
"tfrecord": data_dir["tfrecord"],
|
277 |
"split": special_split_name,
|
278 |
}
|
279 |
)
|
|
|
281 |
return split_generators
|
282 |
|
283 |
|
284 |
+
def _generate_examples(self, features_file, tfrecord, split):
|
285 |
"""This function returns the examples in the raw (text) form."""
|
286 |
_FEAT_MAPPING_FUNCTIONS = {
|
287 |
"answer_choices": lambda x: [choice.decode("utf-8") for choice in x],
|
|
|
295 |
}
|
296 |
|
297 |
key = 0
|
298 |
+
ds = load_cached_task(features_file, tfrecord, split)
|
299 |
for ex in ds.as_numpy_iterator():
|
300 |
ex_dict = {}
|
301 |
for feat_name, feat_value in ex.items():
|