vladsaveliev commited on
Commit
4a3121a
1 Parent(s): 63e7ca3

Add headers to the examples

Browse files
Files changed (3) hide show
  1. .gitignore +1 -0
  2. README.md +16 -6
  3. guitar_tab.py +68 -67
.gitignore CHANGED
@@ -1 +1,2 @@
1
  .DS_Store
 
 
1
  .DS_Store
2
+ .idea
README.md CHANGED
@@ -11,12 +11,8 @@ size_categories:
11
  dataset_info:
12
  - config_name: default
13
  features:
14
- - name: file
15
- dtype: string
16
  - name: text
17
  dtype: string
18
- - name: title
19
- dtype: string
20
  - name: track_name
21
  dtype: string
22
  - name: track_number
@@ -25,18 +21,32 @@ dataset_info:
25
  dtype: string
26
  - name: instrument_number
27
  dtype: int32
 
 
28
  - name: tempo
29
  dtype: int32
30
  - name: tuning
31
  dtype: string
32
  - name: frets
33
  dtype: int32
 
 
 
 
 
 
 
 
 
 
 
 
34
  splits:
35
  - name: train
36
- num_bytes: 951605310
37
  num_examples: 188297
38
  download_size: 75369951
39
- dataset_size: 951605310
40
  - config_name: all
41
  features:
42
  - name: file
 
11
  dataset_info:
12
  - config_name: default
13
  features:
 
 
14
  - name: text
15
  dtype: string
 
 
16
  - name: track_name
17
  dtype: string
18
  - name: track_number
 
21
  dtype: string
22
  - name: instrument_number
23
  dtype: int32
24
+ - name: title
25
+ dtype: string
26
  - name: tempo
27
  dtype: int32
28
  - name: tuning
29
  dtype: string
30
  - name: frets
31
  dtype: int32
32
+ - name: file
33
+ dtype: string
34
+ - name: subtitle
35
+ dtype: string
36
+ - name: artist
37
+ dtype: string
38
+ - name: album
39
+ dtype: string
40
+ - name: words
41
+ dtype: string
42
+ - name: music
43
+ dtype: string
44
  splits:
45
  - name: train
46
+ num_bytes: 955323858
47
  num_examples: 188297
48
  download_size: 75369951
49
+ dataset_size: 955323858
50
  - config_name: all
51
  features:
52
  - name: file
guitar_tab.py CHANGED
@@ -1,10 +1,12 @@
 
1
  import itertools
2
  from pathlib import Path
 
 
3
  import datasets
4
 
5
  datasets.logging.set_verbosity_info()
6
 
7
-
8
  _DESCRIPTION = """\
9
  Dataset of music tablature, in alphaTex (https://alphatab.net/docs/alphatex)
10
  format, converted from Guitar Pro files (gp3, gp4, gp5, which are downloaded
@@ -143,32 +145,37 @@ INSTRUMENT_MAP = {
143
  }
144
 
145
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
146
  class Builder(datasets.GeneratorBasedBuilder):
147
- VERSION = datasets.Version("1.3.1")
148
 
149
  def _info(self):
150
  return datasets.DatasetInfo(
151
  description=_DESCRIPTION,
152
  features=datasets.Features(
153
  {
154
- "file": datasets.Value("string"),
155
- "text": datasets.Value("string"),
156
- "title": datasets.Value("string"),
157
- "track_name": datasets.Value("string"),
158
- "track_number": datasets.Value("int32"),
159
- "instrument_name": datasets.Value("string"),
160
- "instrument_number": datasets.Value("int32"),
161
- "tempo": datasets.Value("int32"),
162
- "tuning": datasets.Value("string"),
163
- "frets": datasets.Value("int32"),
164
- # "subtitle": datasets.Value("string"),
165
- # "artist": datasets.Value("string"),
166
- # "album": datasets.Value("string"),
167
- # "words": datasets.Value("string"),
168
- # "music": datasets.Value("string"),
169
- # "copyright": datasets.Value("string"),
170
- # "tab": datasets.Value("string"),
171
- # "instructions": datasets.Value("string"),
172
  }
173
  ),
174
  )
@@ -194,39 +201,30 @@ class Builder(datasets.GeneratorBasedBuilder):
194
  with open(path) as f:
195
  examples = _parse_examples(f.read())
196
  for example in examples:
197
- example["file"] = path.name
198
- yield f"{idx}-{example['track_number']}", example
199
 
200
 
201
- def _parse_examples(tex):
202
  """
203
  Returns a dictionary for each track
204
  """
205
- tmpl = {
206
- "text": None,
207
- "file": None,
208
- "title": None,
209
- "track_name": None,
210
- "track_number": None,
211
- "instrument_name": None,
212
- "instrument_number": None,
213
- "tempo": None,
214
- "tuning": None,
215
- "frets": None,
216
- }
217
- tracks = [tmpl.copy()]
218
 
219
  def _parse_track_header(k, v):
220
  if k == "instrument":
221
- tracks[-1]["instrument_name"] = INSTRUMENT_MAP.get(int(v), "unknown")
222
- tracks[-1]["instrument_number"] = int(v)
223
  elif k == "tuning":
224
- tracks[-1][k] = v.split(" ")
225
  elif k == "frets":
226
- tracks[-1][k] = int(v)
227
 
228
- lines = (l.strip() for l in tex.splitlines() if l.strip())
229
- for line in itertools.takewhile(lambda l: l != ".", lines):
 
 
230
  assert line.startswith("\\")
231
  line = line.lstrip("\\")
232
  kv = line.split(" ", 1)
@@ -235,44 +233,47 @@ def _parse_examples(tex):
235
  k, v = kv
236
  if k in [
237
  "title",
238
- # "subtitle",
239
- # "artist",
240
- # "album",
241
- # "words",
242
- # "music",
243
- # "copyright",
244
- # "tab",
245
- # "instructions",
246
  ]:
247
- tracks[-1][k] = v.strip('"').strip("'")
248
  elif k == "tempo":
249
- tracks[-1][k] = int(v)
250
  else:
251
  _parse_track_header(k, v)
252
 
253
- track_lines = []
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
254
  for line in lines:
255
  if line.startswith("\\"):
256
- line = line.lstrip("\\")
257
- kv = line.split(" ", 1)
258
- if len(kv) == 2:
259
  k = kv[0]
260
  v = kv[1].strip('"').strip("'")
261
  else:
262
  k = kv[0]
263
  v = None
264
  if k == "track":
265
- if track_lines:
266
- tracks[-1]["text"] = " ".join(track_lines)
267
- track_lines = []
268
- tracks.append(tmpl.copy())
269
- track_number = len(tracks) + 1
270
- tracks[-1]["track_number"] = track_number
271
- tracks[-1]["track_name"] = v if v else f"Track {track_number}"
272
  else:
273
  _parse_track_header(k, v)
274
  else:
275
- track_lines.append(line)
276
- if track_lines:
277
- tracks[-1]["text"] = " ".join(track_lines)
278
- return [t for t in tracks if t["text"]]
 
1
+ import dataclasses
2
  import itertools
3
  from pathlib import Path
4
+ from typing import List
5
+
6
  import datasets
7
 
8
  datasets.logging.set_verbosity_info()
9
 
 
10
  _DESCRIPTION = """\
11
  Dataset of music tablature, in alphaTex (https://alphatab.net/docs/alphatex)
12
  format, converted from Guitar Pro files (gp3, gp4, gp5, which are downloaded
 
145
  }
146
 
147
 
148
+ @dataclasses.dataclass
149
+ class Example:
150
+ text: str
151
+ track_name: str
152
+ track_number: int
153
+ instrument_name: str
154
+ instrument_number: int
155
+ title: str = None
156
+ tempo: int = None
157
+ tuning: str = None
158
+ frets: int = None
159
+ file: str = None
160
+ subtitle: str = None
161
+ artist: str = None
162
+ album: str = None
163
+ words: str = None
164
+ music: str = None
165
+
166
+
167
  class Builder(datasets.GeneratorBasedBuilder):
168
+ VERSION = datasets.Version("1.4.0")
169
 
170
  def _info(self):
171
  return datasets.DatasetInfo(
172
  description=_DESCRIPTION,
173
  features=datasets.Features(
174
  {
175
+ field: datasets.Value("int32")
176
+ if type_ == int
177
+ else datasets.Value("string")
178
+ for field, type_ in Example.__annotations__.items()
 
 
 
 
 
 
 
 
 
 
 
 
 
 
179
  }
180
  ),
181
  )
 
201
  with open(path) as f:
202
  examples = _parse_examples(f.read())
203
  for example in examples:
204
+ example.file = path.name
205
+ yield f"{idx}-{example.track_number}", example.__dict__
206
 
207
 
208
+ def _parse_examples(tex: str) -> List[Example]:
209
  """
210
  Returns a dictionary for each track
211
  """
212
+ examples = []
213
+ data = {}
 
 
 
 
 
 
 
 
 
 
 
214
 
215
  def _parse_track_header(k, v):
216
  if k == "instrument":
217
+ data['instrument_name'] = INSTRUMENT_MAP.get(int(v), "unknown")
218
+ data['instrument_number'] = int(v)
219
  elif k == "tuning":
220
+ data['tuning'] = v.split(" ")
221
  elif k == "frets":
222
+ data['frets'] = int(v)
223
 
224
+ lines = (l_.strip() for l_ in tex.splitlines() if l_.strip())
225
+ header_lines = list(itertools.takewhile(lambda l: l != ".", lines))
226
+
227
+ for line in header_lines:
228
  assert line.startswith("\\")
229
  line = line.lstrip("\\")
230
  kv = line.split(" ", 1)
 
233
  k, v = kv
234
  if k in [
235
  "title",
236
+ "subtitle",
237
+ "artist",
238
+ "album",
239
+ "words",
240
+ "music",
 
 
 
241
  ]:
242
+ data[k] = v.strip('"').strip("'")
243
  elif k == "tempo":
244
+ data['tempo'] = int(v)
245
  else:
246
  _parse_track_header(k, v)
247
 
248
+ track_header_lines = []
249
+ track_tex_lines = []
250
+
251
+ def _add_example():
252
+ nonlocal data, track_header_lines, track_tex_lines
253
+ if track_tex_lines:
254
+ data['track_number'] = (track_number := len(examples) + 1)
255
+ data['track_name'] = v if v else f"Track {track_number}"
256
+ data['text'] = (
257
+ '\n'.join(track_header_lines + [" ".join(track_tex_lines)])
258
+ )
259
+ examples.append(Example(**data))
260
+ data = {}
261
+ track_header_lines = []
262
+ track_tex_lines = []
263
+
264
  for line in lines:
265
  if line.startswith("\\"):
266
+ if len(kv := line.lstrip("\\").split(" ", 1)) == 2:
 
 
267
  k = kv[0]
268
  v = kv[1].strip('"').strip("'")
269
  else:
270
  k = kv[0]
271
  v = None
272
  if k == "track":
273
+ _add_example()
 
 
 
 
 
 
274
  else:
275
  _parse_track_header(k, v)
276
  else:
277
+ track_tex_lines.append(line)
278
+ _add_example()
279
+ return [t for t in examples if t.text]