Upload loaders.py with huggingface_hub
Browse files- loaders.py +97 -27
loaders.py
CHANGED
@@ -1,6 +1,8 @@
|
|
|
|
1 |
import itertools
|
2 |
-
import logging
|
3 |
import os
|
|
|
|
|
4 |
from tempfile import TemporaryDirectory
|
5 |
from typing import Dict, Mapping, Optional, Sequence, Union
|
6 |
|
@@ -8,11 +10,14 @@ import pandas as pd
|
|
8 |
from datasets import load_dataset as hf_load_dataset
|
9 |
from tqdm import tqdm
|
10 |
|
|
|
11 |
from .operator import SourceOperator
|
12 |
from .stream import MultiStream, Stream
|
13 |
|
|
|
14 |
try:
|
15 |
import ibm_boto3
|
|
|
16 |
# from ibm_botocore.client import ClientError
|
17 |
|
18 |
ibm_boto3_available = True
|
@@ -40,31 +45,35 @@ class LoadHF(Loader):
|
|
40 |
Union[str, Sequence[str], Mapping[str, Union[str, Sequence[str]]]]
|
41 |
] = None
|
42 |
streaming: bool = True
|
43 |
-
cached = False
|
44 |
|
45 |
def process(self):
|
46 |
try:
|
47 |
-
|
48 |
-
|
49 |
-
|
50 |
-
|
51 |
-
|
52 |
-
|
53 |
-
|
54 |
-
|
|
|
|
|
55 |
if self.split is not None:
|
56 |
dataset = {self.split: dataset}
|
57 |
except (
|
58 |
NotImplementedError
|
59 |
): # streaming is not supported for zipped files so we load without streaming
|
60 |
-
|
61 |
-
|
62 |
-
|
63 |
-
|
64 |
-
|
65 |
-
|
66 |
-
|
67 |
-
|
|
|
|
|
|
|
68 |
if self.split is None:
|
69 |
for split in dataset.keys():
|
70 |
dataset[split] = dataset[split].to_iterable_dataset()
|
@@ -92,16 +101,55 @@ class LoadCSV(Loader):
|
|
92 |
)
|
93 |
|
94 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
95 |
class LoadFromIBMCloud(Loader):
|
96 |
endpoint_url_env: str
|
97 |
aws_access_key_id_env: str
|
98 |
aws_secret_access_key_env: str
|
99 |
bucket_name: str
|
100 |
data_dir: str = None
|
101 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
102 |
|
103 |
def _download_from_cos(self, cos, bucket_name, item_name, local_file):
|
104 |
-
|
105 |
try:
|
106 |
response = cos.Object(bucket_name, item_name).get()
|
107 |
size = response["ContentLength"]
|
@@ -120,7 +168,7 @@ class LoadFromIBMCloud(Loader):
|
|
120 |
for line in first_lines:
|
121 |
downloaded_file.write(line)
|
122 |
downloaded_file.write(b"\n")
|
123 |
-
|
124 |
f"\nDownload successful limited to {self.loader_limit} lines"
|
125 |
)
|
126 |
return
|
@@ -134,7 +182,7 @@ class LoadFromIBMCloud(Loader):
|
|
134 |
cos.Bucket(bucket_name).download_file(
|
135 |
item_name, local_file, Callback=upload_progress
|
136 |
)
|
137 |
-
|
138 |
except Exception as e:
|
139 |
raise Exception(
|
140 |
f"Unabled to download {item_name} in {bucket_name}", e
|
@@ -145,6 +193,11 @@ class LoadFromIBMCloud(Loader):
|
|
145 |
self.endpoint_url = os.getenv(self.endpoint_url_env)
|
146 |
self.aws_access_key_id = os.getenv(self.aws_access_key_id_env)
|
147 |
self.aws_secret_access_key = os.getenv(self.aws_secret_access_key_env)
|
|
|
|
|
|
|
|
|
|
|
148 |
|
149 |
def verify(self):
|
150 |
super().verify()
|
@@ -166,9 +219,20 @@ class LoadFromIBMCloud(Loader):
|
|
166 |
aws_secret_access_key=self.aws_secret_access_key,
|
167 |
endpoint_url=self.endpoint_url,
|
168 |
)
|
169 |
-
|
170 |
-
|
171 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
172 |
# Build object key based on parameters. Slash character is not
|
173 |
# allowed to be part of object key in IBM COS.
|
174 |
object_key = (
|
@@ -177,8 +241,14 @@ class LoadFromIBMCloud(Loader):
|
|
177 |
else data_file
|
178 |
)
|
179 |
self._download_from_cos(
|
180 |
-
cos, self.bucket_name, object_key,
|
181 |
)
|
182 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
183 |
|
184 |
return MultiStream.from_iterables(dataset)
|
|
|
1 |
+
import importlib
|
2 |
import itertools
|
|
|
3 |
import os
|
4 |
+
import tempfile
|
5 |
+
from pathlib import Path
|
6 |
from tempfile import TemporaryDirectory
|
7 |
from typing import Dict, Mapping, Optional, Sequence, Union
|
8 |
|
|
|
10 |
from datasets import load_dataset as hf_load_dataset
|
11 |
from tqdm import tqdm
|
12 |
|
13 |
+
from .logging_utils import get_logger
|
14 |
from .operator import SourceOperator
|
15 |
from .stream import MultiStream, Stream
|
16 |
|
17 |
+
logger = get_logger()
|
18 |
try:
|
19 |
import ibm_boto3
|
20 |
+
|
21 |
# from ibm_botocore.client import ClientError
|
22 |
|
23 |
ibm_boto3_available = True
|
|
|
45 |
Union[str, Sequence[str], Mapping[str, Union[str, Sequence[str]]]]
|
46 |
] = None
|
47 |
streaming: bool = True
|
|
|
48 |
|
49 |
def process(self):
|
50 |
try:
|
51 |
+
with tempfile.TemporaryDirectory() as dir_to_be_deleted:
|
52 |
+
dataset = hf_load_dataset(
|
53 |
+
self.path,
|
54 |
+
name=self.name,
|
55 |
+
data_dir=self.data_dir,
|
56 |
+
data_files=self.data_files,
|
57 |
+
streaming=self.streaming,
|
58 |
+
cache_dir=None if self.streaming else dir_to_be_deleted,
|
59 |
+
split=self.split,
|
60 |
+
)
|
61 |
if self.split is not None:
|
62 |
dataset = {self.split: dataset}
|
63 |
except (
|
64 |
NotImplementedError
|
65 |
): # streaming is not supported for zipped files so we load without streaming
|
66 |
+
with tempfile.TemporaryDirectory() as dir_to_be_deleted:
|
67 |
+
dataset = hf_load_dataset(
|
68 |
+
self.path,
|
69 |
+
name=self.name,
|
70 |
+
data_dir=self.data_dir,
|
71 |
+
data_files=self.data_files,
|
72 |
+
streaming=False,
|
73 |
+
keep_in_memory=True,
|
74 |
+
cache_dir=dir_to_be_deleted,
|
75 |
+
split=self.split,
|
76 |
+
)
|
77 |
if self.split is None:
|
78 |
for split in dataset.keys():
|
79 |
dataset[split] = dataset[split].to_iterable_dataset()
|
|
|
101 |
)
|
102 |
|
103 |
|
104 |
+
class MissingKaggleCredentialsError(ValueError):
|
105 |
+
pass
|
106 |
+
|
107 |
+
|
108 |
+
# TODO write how to obtain kaggle credentials
|
109 |
+
class LoadFromKaggle(Loader):
|
110 |
+
url: str
|
111 |
+
|
112 |
+
def verify(self):
|
113 |
+
super().verify()
|
114 |
+
if importlib.util.find_spec("opendatasets") is None:
|
115 |
+
raise ImportError(
|
116 |
+
"Please install opendatasets in order to use the LoadFromKaggle loader (using `pip install opendatasets`) "
|
117 |
+
)
|
118 |
+
if not os.path.isfile("kaggle.json"):
|
119 |
+
raise MissingKaggleCredentialsError(
|
120 |
+
"Please obtain kaggle credentials https://christianjmills.com/posts/kaggle-obtain-api-key-tutorial/ and save them to local ./kaggle.json file"
|
121 |
+
)
|
122 |
+
|
123 |
+
def prepare(self):
|
124 |
+
super().prepare()
|
125 |
+
from opendatasets import download
|
126 |
+
|
127 |
+
self.downloader = download
|
128 |
+
|
129 |
+
def process(self):
|
130 |
+
with TemporaryDirectory() as temp_directory:
|
131 |
+
self.downloader(self.url, temp_directory)
|
132 |
+
dataset = hf_load_dataset(temp_directory, streaming=False)
|
133 |
+
|
134 |
+
return MultiStream.from_iterables(dataset)
|
135 |
+
|
136 |
+
|
137 |
class LoadFromIBMCloud(Loader):
|
138 |
endpoint_url_env: str
|
139 |
aws_access_key_id_env: str
|
140 |
aws_secret_access_key_env: str
|
141 |
bucket_name: str
|
142 |
data_dir: str = None
|
143 |
+
|
144 |
+
# Can be either:
|
145 |
+
# 1. a list of file names, the split of each file is determined by the file name pattern
|
146 |
+
# 2. Mapping: split -> file_name, e.g. {"test" : "test.json", "train": "train.json"}
|
147 |
+
# 3. Mapping: split -> file_names, e.g. {"test" : ["test1.json", "test2.json"], "train": ["train.json"]}
|
148 |
+
data_files: Union[Sequence[str], Mapping[str, Union[str, Sequence[str]]]]
|
149 |
+
caching: bool = True
|
150 |
|
151 |
def _download_from_cos(self, cos, bucket_name, item_name, local_file):
|
152 |
+
logger.info(f"Downloading {item_name} from {bucket_name} COS")
|
153 |
try:
|
154 |
response = cos.Object(bucket_name, item_name).get()
|
155 |
size = response["ContentLength"]
|
|
|
168 |
for line in first_lines:
|
169 |
downloaded_file.write(line)
|
170 |
downloaded_file.write(b"\n")
|
171 |
+
logger.info(
|
172 |
f"\nDownload successful limited to {self.loader_limit} lines"
|
173 |
)
|
174 |
return
|
|
|
182 |
cos.Bucket(bucket_name).download_file(
|
183 |
item_name, local_file, Callback=upload_progress
|
184 |
)
|
185 |
+
logger.info("\nDownload Successful")
|
186 |
except Exception as e:
|
187 |
raise Exception(
|
188 |
f"Unabled to download {item_name} in {bucket_name}", e
|
|
|
193 |
self.endpoint_url = os.getenv(self.endpoint_url_env)
|
194 |
self.aws_access_key_id = os.getenv(self.aws_access_key_id_env)
|
195 |
self.aws_secret_access_key = os.getenv(self.aws_secret_access_key_env)
|
196 |
+
root_dir = os.getenv("UNITXT_IBM_COS_CACHE", None) or os.getcwd()
|
197 |
+
self.cache_dir = os.path.join(root_dir, "ibmcos_datasets")
|
198 |
+
|
199 |
+
if not os.path.exists(self.cache_dir):
|
200 |
+
Path(self.cache_dir).mkdir(parents=True, exist_ok=True)
|
201 |
|
202 |
def verify(self):
|
203 |
super().verify()
|
|
|
219 |
aws_secret_access_key=self.aws_secret_access_key,
|
220 |
endpoint_url=self.endpoint_url,
|
221 |
)
|
222 |
+
local_dir = os.path.join(self.cache_dir, self.bucket_name, self.data_dir)
|
223 |
+
if not os.path.exists(local_dir):
|
224 |
+
Path(local_dir).mkdir(parents=True, exist_ok=True)
|
225 |
+
|
226 |
+
if isinstance(self.data_files, Mapping):
|
227 |
+
data_files_names = list(self.data_files.values())
|
228 |
+
if not isinstance(data_files_names[0], str):
|
229 |
+
data_files_names = list(itertools.chain(*data_files_names))
|
230 |
+
else:
|
231 |
+
data_files_names = self.data_files
|
232 |
+
|
233 |
+
for data_file in data_files_names:
|
234 |
+
local_file = os.path.join(local_dir, data_file)
|
235 |
+
if not self.caching or not os.path.exists(local_file):
|
236 |
# Build object key based on parameters. Slash character is not
|
237 |
# allowed to be part of object key in IBM COS.
|
238 |
object_key = (
|
|
|
241 |
else data_file
|
242 |
)
|
243 |
self._download_from_cos(
|
244 |
+
cos, self.bucket_name, object_key, local_dir + "/" + data_file
|
245 |
)
|
246 |
+
|
247 |
+
if isinstance(self.data_files, list):
|
248 |
+
dataset = hf_load_dataset(local_dir, streaming=False)
|
249 |
+
else:
|
250 |
+
dataset = hf_load_dataset(
|
251 |
+
local_dir, streaming=False, data_files=self.data_files
|
252 |
+
)
|
253 |
|
254 |
return MultiStream.from_iterables(dataset)
|