Update irmas.py
Browse files
irmas.py
CHANGED
@@ -102,10 +102,14 @@ class IRMAS(datasets.GeneratorBasedBuilder):
|
|
102 |
# Find all occurrences of the pattern
|
103 |
items = re.findall(pattern, filename)
|
104 |
return items
|
|
|
|
|
|
|
105 |
|
106 |
if split == 'train':
|
107 |
for guid, audio_path in enumerate(audio_filepaths):
|
108 |
labels = extract_bracketed_items(audio_path)
|
|
|
109 |
labels = [label for label in labels if label in INSTRUMENTS]
|
110 |
yield guid, {
|
111 |
"id": str(guid),
|
@@ -121,6 +125,7 @@ class IRMAS(datasets.GeneratorBasedBuilder):
|
|
121 |
with open(audio_path.replace('.wav', '.txt'), 'r') as f:
|
122 |
for line in f:
|
123 |
labels.append(line.strip())
|
|
|
124 |
yield guid, {
|
125 |
"id": str(guid),
|
126 |
"file": audio_path,
|
|
|
102 |
# Find all occurrences of the pattern
|
103 |
items = re.findall(pattern, filename)
|
104 |
return items
|
105 |
+
|
106 |
+
def deduplicate(lst):
|
107 |
+
return list(dict.fromkeys(lst))
|
108 |
|
109 |
if split == 'train':
|
110 |
for guid, audio_path in enumerate(audio_filepaths):
|
111 |
labels = extract_bracketed_items(audio_path)
|
112 |
+
labels = deduplicate(labels)
|
113 |
labels = [label for label in labels if label in INSTRUMENTS]
|
114 |
yield guid, {
|
115 |
"id": str(guid),
|
|
|
125 |
with open(audio_path.replace('.wav', '.txt'), 'r') as f:
|
126 |
for line in f:
|
127 |
labels.append(line.strip())
|
128 |
+
labels = deduplicate(labels)
|
129 |
yield guid, {
|
130 |
"id": str(guid),
|
131 |
"file": audio_path,
|