AtlasUnified commited on
Commit
f24d16b
·
1 Parent(s): 4649fcf

Update atlas-preprocessed-code.py

Browse files
Files changed (1) hide show
  1. atlas-preprocessed-code.py +17 -17
atlas-preprocessed-code.py CHANGED
@@ -6,41 +6,41 @@ import os
6
  logger = datasets.logging.get_logger(__name__)
7
 
8
  _DESCRIPTION = """\
9
- RedPajama is a clean-room, fully open-source implementation of the LLaMa dataset.
10
  """
11
 
12
  # Define a list of programming languages for the subsets
13
- _LANGUAGE_SUBSETS = ["csharp", "cplusplus", "c", "go", "html", "haskell",
14
  "java", "javascript", "jupyter", "kotlin", "php", "perl",
15
  "python", "ruby", "rust", "shell", "swift", "typescript", "v"]
16
 
17
  # Generate URLs for the subsets
18
- _URL_LISTS = {language: f"https://huggingface.co/datasets/AtlasUnified/atlas-preprocessed-code/blob/main/{language}.jsonl"
19
  for language in _LANGUAGE_SUBSETS}
20
 
21
- _URL_BASE = 'https://data.together.xyz/redpajama-data-1T/v1.0.0'
22
 
23
- _DATA_DIR = os.environ.get('RED_PAJAMA_DATA_DIR', None)
24
 
25
- class RedPajama1TConfig(datasets.BuilderConfig):
26
- """BuilderConfig for RedPajama sample."""
27
 
28
  def __init__(self, *args, subsets, **kwargs):
29
- """BuilderConfig for RedPajama.
30
  Args:
31
  **kwargs: keyword arguments forwarded to super.
32
  """
33
- super(RedPajama1TConfig, self).__init__(**kwargs)
34
  self.subsets = subsets
35
 
36
- class RedPajama1T(datasets.GeneratorBasedBuilder):
37
- """RedPajama: Reproducing the LLaMA training dataset of over 1.2 trillion tokens. Version 1.0.0."""
38
 
39
  # Generate builder configs for each language subset
40
- BUILDER_CONFIGS = [RedPajama1TConfig(name=language,
41
  subsets=[language],
42
  version=datasets.Version("1.0.0", ""),
43
- description=f"RedPajama1T {language} subset")
44
  for language in _LANGUAGE_SUBSETS]
45
 
46
  def _info(self):
@@ -50,7 +50,7 @@ class RedPajama1T(datasets.GeneratorBasedBuilder):
50
  {
51
  "text": datasets.Value("string"),
52
  "meta": datasets.Value("string"),
53
- "red_pajama_subset": datasets.Value("string"),
54
  }
55
  ),
56
  supervised_keys=None,
@@ -107,13 +107,13 @@ class RedPajama1T(datasets.GeneratorBasedBuilder):
107
  yield key, {
108
  "text": text,
109
  "meta": json.dumps(data),
110
- "red_pajama_subset": subset,
111
  }
112
  else:
113
  yield key, {
114
  "text": data["text"],
115
  "meta": data["meta"],
116
- "red_pajama_subset": subset,
117
  }
118
  key += 1
119
  except Exception as e:
@@ -122,4 +122,4 @@ class RedPajama1T(datasets.GeneratorBasedBuilder):
122
  print(f'Row: {row}')
123
  traceback.print_exc()
124
 
125
- raise e
 
6
  logger = datasets.logging.get_logger(__name__)
7
 
8
  _DESCRIPTION = """\
9
+ AtlasCode is a clean-room, fully open-source implementation of the LLaMa dataset.
10
  """
11
 
12
  # Define a list of programming languages for the subsets
13
+ _LANGUAGE_SUBSETS = ["c#", "c++", "c", "go", "html", "haskell",
14
  "java", "javascript", "jupyter", "kotlin", "php", "perl",
15
  "python", "ruby", "rust", "shell", "swift", "typescript", "v"]
16
 
17
  # Generate URLs for the subsets
18
+ _URL_LISTS = {language: f"https://huggingface.co/datasets/AtlasUnified/atlas-preprocessed-code/resolve/main/{language}.jsonl"
19
  for language in _LANGUAGE_SUBSETS}
20
 
21
+ _URL_BASE = 'https://huggingface.co/datasets/AtlasUnified/'
22
 
23
+ _DATA_DIR = os.environ.get('ATLAS_CODE_DATA_DIR', None)
24
 
25
+ class AtlasCodeConfig(datasets.BuilderConfig):
26
+ """BuilderConfig for AtlasCode sample."""
27
 
28
  def __init__(self, *args, subsets, **kwargs):
29
+ """BuilderConfig for AtlasCode.
30
  Args:
31
  **kwargs: keyword arguments forwarded to super.
32
  """
33
+ super(AtlasCodeConfig, self).__init__(**kwargs)
34
  self.subsets = subsets
35
 
36
+ class AtlasCode(datasets.GeneratorBasedBuilder):
37
+ """AtlasCode: Reproducing the LLaMA training dataset of over 1.2 trillion tokens. Version 1.0.0."""
38
 
39
  # Generate builder configs for each language subset
40
+ BUILDER_CONFIGS = [AtlasCodeConfig(name=language,
41
  subsets=[language],
42
  version=datasets.Version("1.0.0", ""),
43
+ description=f"AtlasCode {language} subset")
44
  for language in _LANGUAGE_SUBSETS]
45
 
46
  def _info(self):
 
50
  {
51
  "text": datasets.Value("string"),
52
  "meta": datasets.Value("string"),
53
+ "atlas_code_subset": datasets.Value("string"),
54
  }
55
  ),
56
  supervised_keys=None,
 
107
  yield key, {
108
  "text": text,
109
  "meta": json.dumps(data),
110
+ "atlas_code_subset": subset,
111
  }
112
  else:
113
  yield key, {
114
  "text": data["text"],
115
  "meta": data["meta"],
116
+ "atlas_code_subset": subset,
117
  }
118
  key += 1
119
  except Exception as e:
 
122
  print(f'Row: {row}')
123
  traceback.print_exc()
124
 
125
+ raise e