Fixes to incorporate with the latest Lhotse release (#1249)

egs/aidatatang_200zh/ASR/pruned_transducer_stateless2/asr_datamodule.py

@@ -37,7 +37,7 @@ from lhotse.dataset import (
     DynamicBucketingSampler,
     K2SpeechRecognitionDataset,
     PrecomputedFeatures,
-    SingleCutSampler,
+    SimpleCutSampler,
     SpecAugment,
 )
 from lhotse.dataset.input_strategies import OnTheFlyFeatures
@@ -291,8 +291,8 @@ class Aidatatang_200zhAsrDataModule:
                 drop_last=True,
             )
         else:
-            logging.info("Using SingleCutSampler.")
+            logging.info("Using SimpleCutSampler.")
-            train_sampler = SingleCutSampler(
+            train_sampler = SimpleCutSampler(
                 cuts_train,
                 max_duration=self.args.max_duration,
                 shuffle=self.args.shuffle,

egs/aishell/ASR/tdnn_lstm_ctc/asr_datamodule.py

@@ -30,7 +30,7 @@ from lhotse.dataset import (
     DynamicBucketingSampler,
     K2SpeechRecognitionDataset,
     PrecomputedFeatures,
-    SingleCutSampler,
+    SimpleCutSampler,
     SpecAugment,
 )
 from lhotse.dataset.input_strategies import OnTheFlyFeatures
@@ -278,8 +278,8 @@ class AishellAsrDataModule:
                 drop_last=self.args.drop_last,
             )
         else:
-            logging.info("Using SingleCutSampler.")
+            logging.info("Using SimpleCutSampler.")
-            train_sampler = SingleCutSampler(
+            train_sampler = SimpleCutSampler(
                 cuts_train,
                 max_duration=self.args.max_duration,
                 shuffle=self.args.shuffle,

egs/aishell2/ASR/pruned_transducer_stateless5/asr_datamodule.py

@@ -31,7 +31,7 @@ from lhotse.dataset import (  # noqa F401 for PrecomputedFeatures
     DynamicBucketingSampler,
     K2SpeechRecognitionDataset,
     PrecomputedFeatures,
-    SingleCutSampler,
+    SimpleCutSampler,
     SpecAugment,
 )
 from lhotse.dataset.input_strategies import (  # noqa F401 For AudioSamples
@@ -299,8 +299,8 @@ class AiShell2AsrDataModule:
                 drop_last=self.args.drop_last,
             )
         else:
-            logging.info("Using SingleCutSampler.")
+            logging.info("Using SimpleCutSampler.")
-            train_sampler = SingleCutSampler(
+            train_sampler = SimpleCutSampler(
                 cuts_train,
                 max_duration=self.args.max_duration,
                 shuffle=self.args.shuffle,

egs/aishell4/ASR/pruned_transducer_stateless5/asr_datamodule.py

@@ -30,7 +30,7 @@ from lhotse.dataset import (  # noqa F401 for PrecomputedFeatures
     DynamicBucketingSampler,
     K2SpeechRecognitionDataset,
     PrecomputedFeatures,
-    SingleCutSampler,
+    SimpleCutSampler,
     SpecAugment,
 )
 from lhotse.dataset.input_strategies import (  # noqa F401 for AudioSamples
@@ -310,8 +310,8 @@ class Aishell4AsrDataModule:
                 drop_last=self.args.drop_last,
             )
         else:
-            logging.info("Using SingleCutSampler.")
+            logging.info("Using SimpleCutSampler.")
-            train_sampler = SingleCutSampler(
+            train_sampler = SimpleCutSampler(
                 cuts_train,
                 max_duration=self.args.max_duration,
                 shuffle=self.args.shuffle,

egs/alimeeting/ASR/pruned_transducer_stateless2/asr_datamodule.py

@@ -37,7 +37,7 @@ from lhotse.dataset import (
     DynamicBucketingSampler,
     K2SpeechRecognitionDataset,
     PrecomputedFeatures,
-    SingleCutSampler,
+    SimpleCutSampler,
     SpecAugment,
 )
 from lhotse.dataset.input_strategies import OnTheFlyFeatures
@@ -292,8 +292,8 @@ class AlimeetingAsrDataModule:
                 drop_last=True,
             )
         else:
-            logging.info("Using SingleCutSampler.")
+            logging.info("Using SimpleCutSampler.")
-            train_sampler = SingleCutSampler(
+            train_sampler = SimpleCutSampler(
                 cuts_train,
                 max_duration=self.args.max_duration,
                 shuffle=self.args.shuffle,

egs/ami/SURT/dprnn_zipformer/asr_datamodule.py

@@ -257,7 +257,7 @@ class AmiAsrDataModule:
                 drop_last=self.args.drop_last,
             )
         else:
-            logging.info("Using SingleCutSampler.")
+            logging.info("Using SimpleCutSampler.")
             train_sampler = SimpleCutSampler(
                 cuts_train,
                 max_duration=self.args.max_duration,

egs/commonvoice/ASR/pruned_transducer_stateless7/asr_datamodule.py

@@ -30,7 +30,7 @@ from lhotse.dataset import (  # noqa F401 for PrecomputedFeatures
     DynamicBucketingSampler,
     K2SpeechRecognitionDataset,
     PrecomputedFeatures,
-    SingleCutSampler,
+    SimpleCutSampler,
     SpecAugment,
 )
 from lhotse.dataset.input_strategies import (  # noqa F401 For AudioSamples
@@ -311,8 +311,8 @@ class CommonVoiceAsrDataModule:
                 drop_last=self.args.drop_last,
             )
         else:
-            logging.info("Using SingleCutSampler.")
+            logging.info("Using SimpleCutSampler.")
-            train_sampler = SingleCutSampler(
+            train_sampler = SimpleCutSampler(
                 cuts_train,
                 max_duration=self.args.max_duration,
                 shuffle=self.args.shuffle,

egs/csj/ASR/local/utils/asr_datamodule.py

@@ -31,7 +31,7 @@ from lhotse.dataset import (  # noqa F401 for PrecomputedFeatures
     DynamicBucketingSampler,
     K2SpeechRecognitionDataset,
     PrecomputedFeatures,
-    SingleCutSampler,
+    SimpleCutSampler,
     SpecAugment,
 )
 from lhotse.dataset.input_strategies import (  # noqa F401 For AudioSamples
@@ -339,8 +339,8 @@ class CSJAsrDataModule:
                 drop_last=self.args.drop_last,
             )
         else:
-            logging.info("Using SingleCutSampler.")
+            logging.info("Using SimpleCutSampler.")
-            train_sampler = SingleCutSampler(
+            train_sampler = SimpleCutSampler(
                 cuts_train,
                 max_duration=self.args.max_duration,
                 shuffle=self.args.shuffle,

egs/gigaspeech/ASR/conformer_ctc/asr_datamodule.py

@@ -27,7 +27,7 @@ from lhotse.dataset import (
     DynamicBucketingSampler,
     K2SpeechRecognitionDataset,
     PrecomputedFeatures,
-    SingleCutSampler,
+    SimpleCutSampler,
     SpecAugment,
 )
 from lhotse.dataset.input_strategies import OnTheFlyFeatures
@@ -264,8 +264,8 @@ class GigaSpeechAsrDataModule:
                 drop_last=True,
             )
         else:
-            logging.info("Using SingleCutSampler.")
+            logging.info("Using SimpleCutSampler.")
-            train_sampler = SingleCutSampler(
+            train_sampler = SimpleCutSampler(
                 cuts_train,
                 max_duration=self.args.max_duration,
                 shuffle=self.args.shuffle,

egs/gigaspeech/ASR/pruned_transducer_stateless2/asr_datamodule.py

@@ -30,7 +30,7 @@ from lhotse.dataset import (
     DynamicBucketingSampler,
     K2SpeechRecognitionDataset,
     PrecomputedFeatures,
-    SingleCutSampler,
+    SimpleCutSampler,
     SpecAugment,
 )
 from lhotse.dataset.input_strategies import OnTheFlyFeatures
@@ -297,8 +297,8 @@ class GigaSpeechAsrDataModule:
                 drop_last=True,
             )
         else:
-            logging.info("Using SingleCutSampler.")
+            logging.info("Using SimpleCutSampler.")
-            train_sampler = SingleCutSampler(
+            train_sampler = SimpleCutSampler(
                 cuts_train,
                 max_duration=self.args.max_duration,
                 shuffle=self.args.shuffle,

egs/libricss/SURT/dprnn_zipformer/asr_datamodule.py

@@ -259,7 +259,7 @@ class LibriCssAsrDataModule:
                 drop_last=self.args.drop_last,
             )
         else:
-            logging.info("Using SingleCutSampler.")
+            logging.info("Using SimpleCutSampler.")
             train_sampler = SimpleCutSampler(
                 cuts_train,
                 max_duration=self.args.max_duration,

egs/librispeech/ASR/pruned2_knowledge/asr_datamodule.py

@@ -31,7 +31,7 @@ from lhotse.dataset import (
     CutMix,
     K2SpeechRecognitionDataset,
     PrecomputedFeatures,
-    SingleCutSampler,
+    SimpleCutSampler,
     SpecAugment,
 )
 from lhotse.dataset.input_strategies import OnTheFlyFeatures
@@ -290,8 +290,8 @@ class LibriSpeechAsrDataModule:
                 drop_last=True,
             )
         else:
-            logging.info("Using SingleCutSampler.")
+            logging.info("Using SimpleCutSampler.")
-            train_sampler = SingleCutSampler(
+            train_sampler = SimpleCutSampler(
                 cuts_train,
                 max_duration=self.args.max_duration,
                 shuffle=self.args.shuffle,

egs/librispeech/ASR/pruned_transducer_stateless7/gigaspeech.py

@@ -30,7 +30,7 @@ from lhotse.dataset import (
     DynamicBucketingSampler,
     K2SpeechRecognitionDataset,
     PrecomputedFeatures,
-    SingleCutSampler,
+    SimpleCutSampler,
     SpecAugment,
 )
 from lhotse.dataset.input_strategies import OnTheFlyFeatures
@@ -297,8 +297,8 @@ class GigaSpeechAsrDataModule:
                 drop_last=True,
             )
         else:
-            logging.info("Using SingleCutSampler.")
+            logging.info("Using SimpleCutSampler.")
-            train_sampler = SingleCutSampler(
+            train_sampler = SimpleCutSampler(
                 cuts_train,
                 max_duration=self.args.max_duration,
                 shuffle=self.args.shuffle,

egs/librispeech/ASR/tdnn_lstm_ctc/asr_datamodule.py

@@ -31,7 +31,7 @@ from lhotse.dataset import (  # noqa F401 for PrecomputedFeatures
     DynamicBucketingSampler,
     K2SpeechRecognitionDataset,
     PrecomputedFeatures,
-    SingleCutSampler,
+    SimpleCutSampler,
     SpecAugment,
 )
 from lhotse.dataset.input_strategies import (  # noqa F401 For AudioSamples
@@ -314,8 +314,8 @@ class LibriSpeechAsrDataModule:
                 drop_last=self.args.drop_last,
             )
         else:
-            logging.info("Using SingleCutSampler.")
+            logging.info("Using SimpleCutSampler.")
-            train_sampler = SingleCutSampler(
+            train_sampler = SimpleCutSampler(
                 cuts_train,
                 max_duration=self.args.max_duration,
                 shuffle=self.args.shuffle,

egs/mgb2/ASR/conformer_ctc/asr_datamodule.py

@@ -17,7 +17,7 @@ from lhotse.dataset import (
     DynamicBucketingSampler,
     K2SpeechRecognitionDataset,
     PrecomputedFeatures,
-    SingleCutSampler,
+    SimpleCutSampler,
     SpecAugment,
 )
 from lhotse.dataset.input_strategies import OnTheFlyFeatures
@@ -270,8 +270,8 @@ class MGB2AsrDataModule:
                 drop_last=self.args.drop_last,
             )
         else:
-            logging.info("Using SingleCutSampler.")
+            logging.info("Using SimpleCutSampler.")
-            train_sampler = SingleCutSampler(
+            train_sampler = SimpleCutSampler(
                 cuts_train,
                 max_duration=self.args.max_duration,
                 shuffle=self.args.shuffle,

egs/multi_zh-hans/ASR/zipformer/asr_datamodule.py

@@ -31,7 +31,7 @@ from lhotse.dataset import (  # noqa F401 for PrecomputedFeatures
     DynamicBucketingSampler,
     K2SpeechRecognitionDataset,
     PrecomputedFeatures,
-    SingleCutSampler,
+    SimpleCutSampler,
     SpecAugment,
 )
 from lhotse.dataset.input_strategies import (  # noqa F401 For AudioSamples
@@ -300,8 +300,8 @@ class AsrDataModule:
                 drop_last=self.args.drop_last,
             )
         else:
-            logging.info("Using SingleCutSampler.")
+            logging.info("Using SimpleCutSampler.")
-            train_sampler = SingleCutSampler(
+            train_sampler = SimpleCutSampler(
                 cuts_train,
                 max_duration=self.args.max_duration,
                 shuffle=self.args.shuffle,

egs/tal_csasr/ASR/pruned_transducer_stateless5/asr_datamodule.py

@@ -30,7 +30,7 @@ from lhotse.dataset import (  # noqa F401 for PrecomputedFeatures
     DynamicBucketingSampler,
     K2SpeechRecognitionDataset,
     PrecomputedFeatures,
-    SingleCutSampler,
+    SimpleCutSampler,
     SpecAugment,
 )
 from lhotse.dataset.input_strategies import (  # noqa F401 for AudioSamples
@@ -311,8 +311,8 @@ class TAL_CSASRAsrDataModule:
                 drop_last=self.args.drop_last,
             )
         else:
-            logging.info("Using SingleCutSampler.")
+            logging.info("Using SimpleCutSampler.")
-            train_sampler = SingleCutSampler(
+            train_sampler = SimpleCutSampler(
                 cuts_train,
                 max_duration=self.args.max_duration,
                 shuffle=self.args.shuffle,

egs/tedlium3/ASR/transducer_stateless/asr_datamodule.py

@@ -28,7 +28,7 @@ from lhotse.dataset import (
     CutMix,
     DynamicBucketingSampler,
     K2SpeechRecognitionDataset,
-    SingleCutSampler,
+    SimpleCutSampler,
     SpecAugment,
 )
 from lhotse.dataset.input_strategies import OnTheFlyFeatures
@@ -259,8 +259,8 @@ class TedLiumAsrDataModule:
                 drop_last=True,
             )
         else:
-            logging.info("Using SingleCutSampler.")
+            logging.info("Using SimpleCutSampler.")
-            train_sampler = SingleCutSampler(
+            train_sampler = SimpleCutSampler(
                 cuts_train,
                 max_duration=self.args.max_duration,
                 shuffle=self.args.shuffle,
@@ -282,7 +282,6 @@ class TedLiumAsrDataModule:
         return train_dl
 
     def valid_dataloaders(self, cuts_valid: CutSet) -> DataLoader:
-
         transforms = []
         if self.args.concatenate_cuts:
             transforms = [
@@ -322,7 +321,6 @@ class TedLiumAsrDataModule:
         return valid_dl
 
     def test_dataloaders(self, cuts_test: CutSet) -> DataLoader:
-
         logging.debug("About to create test dataset")
         if self.args.on_the_fly_feats:
             test = K2SpeechRecognitionDataset(

egs/timit/ASR/tdnn_lstm_ctc/asr_datamodule.py

@@ -30,7 +30,7 @@ from lhotse.dataset import (
     DynamicBucketingSampler,
     K2SpeechRecognitionDataset,
     PrecomputedFeatures,
-    SingleCutSampler,
+    SimpleCutSampler,
     SpecAugment,
 )
 from lhotse.dataset.input_strategies import OnTheFlyFeatures
@@ -225,8 +225,8 @@ class TimitAsrDataModule(DataModule):
                 drop_last=True,
             )
         else:
-            logging.info("Using SingleCutSampler.")
+            logging.info("Using SimpleCutSampler.")
-            train_sampler = SingleCutSampler(
+            train_sampler = SimpleCutSampler(
                 cuts_train,
                 max_duration=self.args.max_duration,
                 shuffle=self.args.shuffle,
@@ -267,7 +267,7 @@ class TimitAsrDataModule(DataModule):
                 cut_transforms=transforms,
                 return_cuts=self.args.return_cuts,
             )
-        valid_sampler = SingleCutSampler(
+        valid_sampler = SimpleCutSampler(
             cuts_valid,
             max_duration=self.args.max_duration,
             shuffle=False,
@@ -298,7 +298,7 @@ class TimitAsrDataModule(DataModule):
                 else PrecomputedFeatures(),
                 return_cuts=self.args.return_cuts,
             )
-            sampler = SingleCutSampler(cuts_test, max_duration=self.args.max_duration)
+            sampler = SimpleCutSampler(cuts_test, max_duration=self.args.max_duration)
             logging.debug("About to create test dataloader")
             test_dl = DataLoader(test, batch_size=None, sampler=sampler, num_workers=1)
             test_loaders.append(test_dl)

egs/wenetspeech/ASR/pruned_transducer_stateless2/asr_datamodule.py

@@ -37,7 +37,7 @@ from lhotse.dataset import (
     DynamicBucketingSampler,
     K2SpeechRecognitionDataset,
     PrecomputedFeatures,
-    SingleCutSampler,
+    SimpleCutSampler,
     SpecAugment,
 )
 from lhotse.dataset.input_strategies import OnTheFlyFeatures
@@ -296,8 +296,8 @@ class WenetSpeechAsrDataModule:
                 drop_last=True,
             )
         else:
-            logging.info("Using SingleCutSampler.")
+            logging.info("Using SimpleCutSampler.")
-            train_sampler = SingleCutSampler(
+            train_sampler = SimpleCutSampler(
                 cuts_train,
                 max_duration=self.args.max_duration,
                 shuffle=self.args.shuffle,

egs/xbmu_amdo31/ASR/pruned_transducer_stateless5/asr_datamodule.py

@@ -32,7 +32,7 @@ from lhotse.dataset import (
     DynamicBucketingSampler,
     K2SpeechRecognitionDataset,
     PrecomputedFeatures,
-    SingleCutSampler,
+    SimpleCutSampler,
     SpecAugment,
 )
 from lhotse.dataset.input_strategies import AudioSamples  # noqa F401 For AudioSamples
@@ -299,8 +299,8 @@ class Xbmu_AmdoAsrDataModule:
                 drop_last=self.args.drop_last,
             )
         else:
-            logging.info("Using SingleCutSampler.")
+            logging.info("Using SimpleCutSampler.")
-            train_sampler = SingleCutSampler(
+            train_sampler = SimpleCutSampler(
                 cuts_train,
                 max_duration=self.args.max_duration,
                 shuffle=self.args.shuffle,

egs/yesno/ASR/tdnn/asr_datamodule.py

@@ -26,7 +26,7 @@ from lhotse.dataset import (
     DynamicBucketingSampler,
     K2SpeechRecognitionDataset,
     PrecomputedFeatures,
-    SingleCutSampler,
+    SimpleCutSampler,
 )
 from lhotse.dataset.input_strategies import OnTheFlyFeatures
 from torch.utils.data import DataLoader
@@ -196,8 +196,8 @@ class YesNoAsrDataModule(DataModule):
                 drop_last=True,
             )
         else:
-            logging.info("Using SingleCutSampler.")
+            logging.info("Using SimpleCutSampler.")
-            train_sampler = SingleCutSampler(
+            train_sampler = SimpleCutSampler(
                 cuts_train,
                 max_duration=self.args.max_duration,
                 shuffle=self.args.shuffle,

requirements-ci.txt

@@ -10,6 +10,7 @@ graphviz==0.19.1
 
 -f https://download.pytorch.org/whl/cpu/torch_stable.html torch==1.13.1+cpu
 -f https://download.pytorch.org/whl/cpu/torch_stable.html torchaudio==0.13.1+cpu
+six
 
 -f https://k2-fsa.org/nightly/ k2==1.23.4.dev20230319+cpu.torch1.13.1
 

test/test_ali.py

@@ -26,7 +26,7 @@
 from pathlib import Path
 
 from lhotse import CutSet, load_manifest
-from lhotse.dataset import K2SpeechRecognitionDataset, SingleCutSampler
+from lhotse.dataset import K2SpeechRecognitionDataset, SimpleCutSampler
 from lhotse.dataset.collation import collate_custom_field
 from torch.utils.data import DataLoader
 
@@ -44,7 +44,7 @@ def get_dataloader():
     cuts = load_manifest(cuts_json)
     print(cuts[0])
     cuts = cuts.with_features_path_prefix(egs_dir)
-    sampler = SingleCutSampler(
+    sampler = SimpleCutSampler(
         cuts,
         max_duration=10,
         shuffle=False,