pianistprogrammer commited on
Commit
ec015f3
·
1 Parent(s): b577f77

Refactor Meter2800 dataset configuration and example generation logic

Browse files
Files changed (1) hide show
  1. meter2800.py +53 -18
meter2800.py CHANGED
@@ -23,20 +23,44 @@ It is split into training, validation, and test sets, each available in two clas
23
  _HOMEPAGE = "https://huggingface.co/datasets/pianistprogrammer/Meter2800"
24
  _LICENSE = "mit"
25
 
26
- # Define the labels used in the "label" and "meter"/"alt_meter" columns
27
- LABELS_4 = ["three", "four", "five" , "seven"]
28
- LABELS_2 = ["three", "four"] # Example if using a binary grouping
 
 
 
 
 
 
 
 
 
29
 
30
  class Meter2800(datasets.GeneratorBasedBuilder):
 
 
31
  BUILDER_CONFIGS = [
32
- datasets.BuilderConfig(name="4_classes", version=datasets.Version("1.0.0"),
33
- description="4-class meter classification"),
34
- datasets.BuilderConfig(name="2_classes", version=datasets.Version("1.0.0"),
35
- description="2-class meter classification"),
 
 
 
 
36
  ]
 
 
37
 
38
  def _info(self):
39
- label_names = LABELS_4 if self.config.name == "4_classes" else LABELS_2
 
 
 
 
 
 
 
40
  return datasets.DatasetInfo(
41
  description=_DESCRIPTION,
42
  features=datasets.Features({
@@ -53,34 +77,45 @@ class Meter2800(datasets.GeneratorBasedBuilder):
53
  )
54
 
55
  def _split_generators(self, dl_manager):
56
- root = Path(__file__).parent
57
- suffix = "4_classes" if self.config.name == "4_classes" else "2_classes"
58
-
59
  return [
60
  datasets.SplitGenerator(
61
  name=datasets.Split.TRAIN,
62
- gen_kwargs={"csv_file": root / f"data_train_{suffix}.csv"},
 
 
 
63
  ),
64
  datasets.SplitGenerator(
65
  name=datasets.Split.VALIDATION,
66
- gen_kwargs={"csv_file": root / f"data_val_{suffix}.csv"},
 
 
 
67
  ),
68
  datasets.SplitGenerator(
69
  name=datasets.Split.TEST,
70
- gen_kwargs={"csv_file": root / f"data_test_{suffix}.csv"},
 
 
 
71
  ),
72
  ]
73
 
74
- def _generate_examples(self, csv_file):
75
  df = pd.read_csv(csv_file)
76
  df = df.dropna(subset=["filename", "label", "meter"]).reset_index(drop=True)
77
 
78
  for idx, row in df.iterrows():
79
- audio_path = Path(__file__).parent / row["filename"].lstrip("/")
 
 
80
  yield idx, {
81
  "filename": row["filename"],
82
- "audio": str(audio_path),
83
  "label": row["label"],
84
  "meter": str(row["meter"]),
85
  "alt_meter": str(row.get("alt_meter", row["meter"])),
86
- }
 
23
  _HOMEPAGE = "https://huggingface.co/datasets/pianistprogrammer/Meter2800"
24
  _LICENSE = "mit"
25
 
26
+ # Define the labels - adjust these based on your actual data
27
+ LABELS_4 = ["three", "four", "five", "seven"]
28
+ LABELS_2 = ["simple", "complex"] # or whatever your 2-class grouping actually is
29
+
30
+ class Meter2800Config(datasets.BuilderConfig):
31
+ """BuilderConfig for Meter2800."""
32
+ def __init__(self, name, **kwargs):
33
+ super(Meter2800Config, self).__init__(
34
+ name=name,
35
+ version=datasets.Version("1.0.0"),
36
+ **kwargs
37
+ )
38
 
39
  class Meter2800(datasets.GeneratorBasedBuilder):
40
+ """Meter2800 dataset."""
41
+
42
  BUILDER_CONFIGS = [
43
+ Meter2800Config(
44
+ name="4_classes",
45
+ description="4-class meter classification"
46
+ ),
47
+ Meter2800Config(
48
+ name="2_classes",
49
+ description="2-class meter classification"
50
+ ),
51
  ]
52
+
53
+ DEFAULT_CONFIG_NAME = "4_classes"
54
 
55
  def _info(self):
56
+ if self.config.name == "4_classes":
57
+ label_names = LABELS_4
58
+ elif self.config.name == "2_classes":
59
+ label_names = LABELS_2
60
+ else:
61
+ # Fallback - shouldn't happen with proper configs
62
+ label_names = LABELS_4
63
+
64
  return datasets.DatasetInfo(
65
  description=_DESCRIPTION,
66
  features=datasets.Features({
 
77
  )
78
 
79
  def _split_generators(self, dl_manager):
80
+ # Get the data directory
81
+ data_dir = dl_manager.download_and_extract("")
82
+
83
  return [
84
  datasets.SplitGenerator(
85
  name=datasets.Split.TRAIN,
86
+ gen_kwargs={
87
+ "csv_file": f"{data_dir}/data_train_{self.config.name}.csv",
88
+ "data_dir": data_dir
89
+ },
90
  ),
91
  datasets.SplitGenerator(
92
  name=datasets.Split.VALIDATION,
93
+ gen_kwargs={
94
+ "csv_file": f"{data_dir}/data_val_{self.config.name}.csv",
95
+ "data_dir": data_dir
96
+ },
97
  ),
98
  datasets.SplitGenerator(
99
  name=datasets.Split.TEST,
100
+ gen_kwargs={
101
+ "csv_file": f"{data_dir}/data_test_{self.config.name}.csv",
102
+ "data_dir": data_dir
103
+ },
104
  ),
105
  ]
106
 
107
+ def _generate_examples(self, csv_file, data_dir):
108
  df = pd.read_csv(csv_file)
109
  df = df.dropna(subset=["filename", "label", "meter"]).reset_index(drop=True)
110
 
111
  for idx, row in df.iterrows():
112
+ # Construct the full audio path
113
+ audio_path = f"{data_dir}/{row['filename']}"
114
+
115
  yield idx, {
116
  "filename": row["filename"],
117
+ "audio": audio_path,
118
  "label": row["label"],
119
  "meter": str(row["meter"]),
120
  "alt_meter": str(row.get("alt_meter", row["meter"])),
121
+ }