Dan Fu commited on
Commit
4e62b3b
1 Parent(s): 0d59dbd

Update loader script for local files

Browse files
Files changed (2) hide show
  1. README.md +8 -0
  2. RedPajama-Data-1T.py +68 -6
README.md CHANGED
@@ -25,6 +25,14 @@ while read line; do
25
  done < urls.txt
26
  ```
27
 
 
 
 
 
 
 
 
 
28
 
29
  A smaller 1B-token sample of the dataset can be found [here](https://huggingface.co/datasets/togethercomputer/RedPajama-Data-1T-Sample).
30
 
 
25
  done < urls.txt
26
  ```
27
 
28
+ After downloading the files, you can load the dataset from disk by setting the `RED_PAJAMA_DATA_DIR` environment variable to the directory containing the files:
29
+
30
+ ```python
31
+ import os
32
+ from datasets import load_dataset
33
+ os.environ["RED_PAJAMA_DATA_DIR"] = "/path/to/download"
34
+ ds = load_dataset("togethercomputer/RedPajama-Data-1T")
35
+ ```
36
 
37
  A smaller 1B-token sample of the dataset can be found [here](https://huggingface.co/datasets/togethercomputer/RedPajama-Data-1T-Sample).
38
 
RedPajama-Data-1T.py CHANGED
@@ -20,6 +20,7 @@ import json
20
 
21
  import datasets
22
  import traceback
 
23
 
24
  logger = datasets.logging.get_logger(__name__)
25
 
@@ -37,7 +38,9 @@ _URL_LISTS = {
37
  "stackexchange": "urls/stackexchange.txt",
38
  "wikipedia": "urls/wikipedia.txt",
39
  }
 
40
 
 
41
 
42
  class RedPajama1TConfig(datasets.BuilderConfig):
43
  """BuilderConfig for RedPajama sample."""
@@ -48,7 +51,6 @@ class RedPajama1TConfig(datasets.BuilderConfig):
48
  **kwargs: keyword arguments forwarded to super.
49
  """
50
  super(RedPajama1TConfig, self).__init__(**kwargs)
51
-
52
  self.subsets = subsets
53
 
54
 
@@ -57,10 +59,59 @@ class RedPajama1T(datasets.GeneratorBasedBuilder):
57
 
58
  BUILDER_CONFIGS = [
59
  RedPajama1TConfig(
 
60
  subsets = list(_URL_LISTS.keys()),
61
- name="plain_text",
62
  version=datasets.Version("1.0.0", ""),
63
- description="Plain text",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
64
  ),
65
  ]
66
 
@@ -86,9 +137,20 @@ class RedPajama1T(datasets.GeneratorBasedBuilder):
86
 
87
  for subset, url_list in url_lists.items():
88
  with open(url_list, encoding="utf-8") as f:
89
- urls[subset] = [line.strip() for line in f][:1]
90
-
91
- downloaded_files = dl_manager.download(urls)
 
 
 
 
 
 
 
 
 
 
 
92
 
93
  return [
94
  datasets.SplitGenerator(
 
20
 
21
  import datasets
22
  import traceback
23
+ import os
24
 
25
  logger = datasets.logging.get_logger(__name__)
26
 
 
38
  "stackexchange": "urls/stackexchange.txt",
39
  "wikipedia": "urls/wikipedia.txt",
40
  }
41
+ _URL_BASE = 'https://data.together.xyz/redpajama-data-1T/v1.0.0'
42
 
43
+ _DATA_DIR = os.environ.get('RED_PAJAMA_DATA_DIR', None)
44
 
45
  class RedPajama1TConfig(datasets.BuilderConfig):
46
  """BuilderConfig for RedPajama sample."""
 
51
  **kwargs: keyword arguments forwarded to super.
52
  """
53
  super(RedPajama1TConfig, self).__init__(**kwargs)
 
54
  self.subsets = subsets
55
 
56
 
 
59
 
60
  BUILDER_CONFIGS = [
61
  RedPajama1TConfig(
62
+ name = 'default',
63
  subsets = list(_URL_LISTS.keys()),
 
64
  version=datasets.Version("1.0.0", ""),
65
+ description="RedPajama1T",
66
+ ),
67
+
68
+ RedPajama1TConfig(
69
+ name = 'arxiv',
70
+ subsets = 'arxiv',
71
+ version=datasets.Version("1.0.0", ""),
72
+ description="RedPajama1T arxiv subset",
73
+ ),
74
+
75
+ RedPajama1TConfig(
76
+ name = 'book',
77
+ subsets = 'book',
78
+ version=datasets.Version("1.0.0", ""),
79
+ description="RedPajama1T book subset",
80
+ ),
81
+
82
+ RedPajama1TConfig(
83
+ name = 'c4',
84
+ subsets = 'c4',
85
+ version=datasets.Version("1.0.0", ""),
86
+ description="RedPajama1T c4 subset",
87
+ ),
88
+
89
+ RedPajama1TConfig(
90
+ name = 'common_crawl',
91
+ subsets = 'common_crawl',
92
+ version=datasets.Version("1.0.0", ""),
93
+ description="RedPajama1T common crawl subset",
94
+ ),
95
+
96
+ RedPajama1TConfig(
97
+ name = 'github',
98
+ subsets = 'github',
99
+ version=datasets.Version("1.0.0", ""),
100
+ description="RedPajama1T github subset",
101
+ ),
102
+
103
+ RedPajama1TConfig(
104
+ name = 'stackexchange',
105
+ subsets = 'stackexchange',
106
+ version=datasets.Version("1.0.0", ""),
107
+ description="RedPajama1T stackexchange subset",
108
+ ),
109
+
110
+ RedPajama1TConfig(
111
+ name = 'wikipedia',
112
+ subsets = 'wikipedia',
113
+ version=datasets.Version("1.0.0", ""),
114
+ description="RedPajama1T wikipedia subset",
115
  ),
116
  ]
117
 
 
137
 
138
  for subset, url_list in url_lists.items():
139
  with open(url_list, encoding="utf-8") as f:
140
+ urls[subset] = [line.strip() for line in f]
141
+
142
+ if _DATA_DIR is not None:
143
+ print(f'Reading data from {_DATA_DIR}')
144
+ url_prefix_slashes = len(_URL_BASE.split('/'))
145
+ downloaded_files = {
146
+ subset: [
147
+ os.path.join(_DATA_DIR, *url.split('/')[url_prefix_slashes:])
148
+ for url in url_list
149
+ ]
150
+ for subset, url_list in urls.items()
151
+ }
152
+ else:
153
+ downloaded_files = dl_manager.download(urls)
154
 
155
  return [
156
  datasets.SplitGenerator(