Commit ·
ca1e14d
1
Parent(s): ecad711
fix labels downloading
Browse files- transcribed_calls.py +7 -4
transcribed_calls.py
CHANGED
|
@@ -41,7 +41,8 @@ class TranscribedCalls(datasets.GeneratorBasedBuilder):
|
|
| 41 |
|
| 42 |
def _split_generators(self, dl_manager):
|
| 43 |
metadata_path = dl_manager.download("metadata.csv")
|
| 44 |
-
|
|
|
|
| 45 |
|
| 46 |
return [
|
| 47 |
datasets.SplitGenerator(
|
|
@@ -49,6 +50,7 @@ class TranscribedCalls(datasets.GeneratorBasedBuilder):
|
|
| 49 |
# These kwargs will be passed to _generate_examples
|
| 50 |
gen_kwargs={
|
| 51 |
"metadata_path": metadata_path,
|
|
|
|
| 52 |
"audio_files": dl_manager.iter_archive(archive),
|
| 53 |
"split": "train",
|
| 54 |
},
|
|
@@ -58,6 +60,7 @@ class TranscribedCalls(datasets.GeneratorBasedBuilder):
|
|
| 58 |
# These kwargs will be passed to _generate_examples
|
| 59 |
gen_kwargs={
|
| 60 |
"metadata_path": metadata_path,
|
|
|
|
| 61 |
"audio_files": dl_manager.iter_archive(archive),
|
| 62 |
"split": "test"
|
| 63 |
},
|
|
@@ -65,7 +68,7 @@ class TranscribedCalls(datasets.GeneratorBasedBuilder):
|
|
| 65 |
]
|
| 66 |
|
| 67 |
# method parameters are unpacked from `gen_kwargs` as given in `_split_generators`
|
| 68 |
-
def _generate_examples(self, metadata_path, audio_files, split):
|
| 69 |
# TODO: This method handles input defined in _split_generators to yield (key, example) tuples from the dataset.
|
| 70 |
# The `key` is for legacy reasons (tfds) and is not important in itself, but must be unique for each example.
|
| 71 |
print(f"generating examples for split {split}")
|
|
@@ -77,12 +80,12 @@ class TranscribedCalls(datasets.GeneratorBasedBuilder):
|
|
| 77 |
if split != split:
|
| 78 |
continue
|
| 79 |
|
| 80 |
-
words_path = os.path.join(
|
| 81 |
words = ''
|
| 82 |
with open(words_path, "r") as f:
|
| 83 |
words = f.read()
|
| 84 |
|
| 85 |
-
phonemes_path = os.path.join(
|
| 86 |
phonemes = ''
|
| 87 |
with open(phonemes_path, "r") as f:
|
| 88 |
phonemes = f.read()
|
|
|
|
| 41 |
|
| 42 |
def _split_generators(self, dl_manager):
|
| 43 |
metadata_path = dl_manager.download("metadata.csv")
|
| 44 |
+
data_dir = dl_manager.download("data/audio.tar.gz")
|
| 45 |
+
archive = os.path.join(data_dir, "audio.tar.gz")
|
| 46 |
|
| 47 |
return [
|
| 48 |
datasets.SplitGenerator(
|
|
|
|
| 50 |
# These kwargs will be passed to _generate_examples
|
| 51 |
gen_kwargs={
|
| 52 |
"metadata_path": metadata_path,
|
| 53 |
+
"data_dir": data_dir,
|
| 54 |
"audio_files": dl_manager.iter_archive(archive),
|
| 55 |
"split": "train",
|
| 56 |
},
|
|
|
|
| 60 |
# These kwargs will be passed to _generate_examples
|
| 61 |
gen_kwargs={
|
| 62 |
"metadata_path": metadata_path,
|
| 63 |
+
"data_dir": data_dir,
|
| 64 |
"audio_files": dl_manager.iter_archive(archive),
|
| 65 |
"split": "test"
|
| 66 |
},
|
|
|
|
| 68 |
]
|
| 69 |
|
| 70 |
# method parameters are unpacked from `gen_kwargs` as given in `_split_generators`
|
| 71 |
+
def _generate_examples(self, metadata_path, data_dir, audio_files, split):
|
| 72 |
# TODO: This method handles input defined in _split_generators to yield (key, example) tuples from the dataset.
|
| 73 |
# The `key` is for legacy reasons (tfds) and is not important in itself, but must be unique for each example.
|
| 74 |
print(f"generating examples for split {split}")
|
|
|
|
| 80 |
if split != split:
|
| 81 |
continue
|
| 82 |
|
| 83 |
+
words_path = os.path.join(data_dir, "words", id_ + ".txt")
|
| 84 |
words = ''
|
| 85 |
with open(words_path, "r") as f:
|
| 86 |
words = f.read()
|
| 87 |
|
| 88 |
+
phonemes_path = os.path.join(data_dir, "phonemes", id_ + ".txt")
|
| 89 |
phonemes = ''
|
| 90 |
with open(phonemes_path, "r") as f:
|
| 91 |
phonemes = f.read()
|