Added the popular GTZAN dataset:

docs/source/datasets.rst

@@ -33,6 +33,14 @@ COMMONVOICE
   :special-members:
 
 
+GTZAN
+~~~~~
+
+.. autoclass:: GTZAN
+  :members: __getitem__
+  :special-members:
+
+
 LIBRISPEECH
 ~~~~~~~~~~~
 

test/test_datasets.py

@@ -7,6 +7,7 @@
 from torchaudio.datasets.vctk import VCTK
 from torchaudio.datasets.yesno import YESNO
 from torchaudio.datasets.ljspeech import LJSPEECH
+from torchaudio.datasets.gtzan import GTZAN
 
 from . import common_utils
 
@@ -55,6 +56,10 @@ def test_speechcommands(self):
         data = SPEECHCOMMANDS(self.path)
         data[0]
 
+    def test_gtzan(self):
+        data = GTZAN(self.path)
+        data[0]
+
 
 if __name__ == "__main__":
     unittest.main()

torchaudio/datasets/__init__.py

@@ -3,6 +3,7 @@
 from .speechcommands import SPEECHCOMMANDS
 from .utils import bg_iterator, diskcache_iterator
 from .vctk import VCTK
+from .gtzan import GTZAN
 from .yesno import YESNO
 from .ljspeech import LJSPEECH
 
@@ -13,6 +14,7 @@
     "VCTK",
     "YESNO",
     "LJSPEECH",
+    "GTZAN",
     "diskcache_iterator",
     "bg_iterator",
 )