From 736a60fb4863de694c36ea4b2391da9b83f973c0 Mon Sep 17 00:00:00 2001 From: jinzr Date: Tue, 31 Oct 2023 10:14:52 +0800 Subject: [PATCH] incorporate https://github.com/k2-fsa/icefall/pull/1301 --- egs/gigaspeech/ASR/zipformer/asr_datamodule.py | 2 +- egs/librispeech/ASR/tiny_transducer_ctc/asr_datamodule.py | 8 ++++---- 2 files changed, 5 insertions(+), 5 deletions(-) diff --git a/egs/gigaspeech/ASR/zipformer/asr_datamodule.py b/egs/gigaspeech/ASR/zipformer/asr_datamodule.py index c4472ed23..6adfdbfbb 100644 --- a/egs/gigaspeech/ASR/zipformer/asr_datamodule.py +++ b/egs/gigaspeech/ASR/zipformer/asr_datamodule.py @@ -233,7 +233,7 @@ class GigaSpeechAsrDataModule: logging.info("About to get Musan cuts") cuts_musan = load_manifest(self.args.manifest_dir / "musan_cuts.jsonl.gz") transforms.append( - CutMix(cuts=cuts_musan, prob=0.5, snr=(10, 20), preserve_id=True) + CutMix(cuts=cuts_musan, p=0.5, snr=(10, 20), preserve_id=True) ) else: logging.info("Disable MUSAN") diff --git a/egs/librispeech/ASR/tiny_transducer_ctc/asr_datamodule.py b/egs/librispeech/ASR/tiny_transducer_ctc/asr_datamodule.py index 8facb6dba..3acd22ae4 100644 --- a/egs/librispeech/ASR/tiny_transducer_ctc/asr_datamodule.py +++ b/egs/librispeech/ASR/tiny_transducer_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 @@ -225,7 +225,7 @@ class LibriSpeechAsrDataModule: logging.info("About to get Musan cuts") cuts_musan = load_manifest(self.args.manifest_dir / "musan_cuts.jsonl.gz") transforms.append( - CutMix(cuts=cuts_musan, prob=0.5, snr=(10, 20), preserve_id=True) + CutMix(cuts=cuts_musan, p=0.5, snr=(10, 20), preserve_id=True) ) else: logging.info("Disable MUSAN") @@ -307,8 +307,8 @@ class LibriSpeechAsrDataModule: drop_last=self.args.drop_last, ) else: - logging.info("Using SingleCutSampler.") - train_sampler = SingleCutSampler( + logging.info("Using SimpleCutSampler.") + train_sampler = SimpleCutSampler( cuts_train, max_duration=self.args.max_duration, shuffle=self.args.shuffle,