diff --git a/configs/ljs_base.json b/configs/ljs_base.json index c689b54a..810f89e7 100644 --- a/configs/ljs_base.json +++ b/configs/ljs_base.json @@ -20,6 +20,7 @@ "training_files":"filelists/ljs_audio_text_train_filelist.txt.cleaned", "validation_files":"filelists/ljs_audio_text_val_filelist.txt.cleaned", "text_cleaners":["english_cleaners2"], + "audio_boundaries": [32,300,400,500,600,700,800,900,1000], "max_wav_value": 32768.0, "sampling_rate": 22050, "filter_length": 1024, diff --git a/configs/ljs_nosdp.json b/configs/ljs_nosdp.json index 7d273775..b10e8694 100644 --- a/configs/ljs_nosdp.json +++ b/configs/ljs_nosdp.json @@ -21,6 +21,7 @@ "validation_files":"filelists/ljs_audio_text_val_filelist.txt.cleaned", "text_cleaners":["english_cleaners2"], "max_wav_value": 32768.0, + "audio_boundaries": [32,300,400,500,600,700,800,900,1000], "sampling_rate": 22050, "filter_length": 1024, "hop_length": 256, diff --git a/configs/vctk_base.json b/configs/vctk_base.json index 3a656f4f..96c0381a 100644 --- a/configs/vctk_base.json +++ b/configs/vctk_base.json @@ -21,6 +21,7 @@ "validation_files":"filelists/vctk_audio_sid_text_val_filelist.txt.cleaned", "text_cleaners":["english_cleaners2"], "max_wav_value": 32768.0, + "audio_boundaries": [32,300,400,500,600,700,800,900,1000], "sampling_rate": 22050, "filter_length": 1024, "hop_length": 256, diff --git a/train.py b/train.py index 703d30cf..e5b9dfe8 100644 --- a/train.py +++ b/train.py @@ -67,7 +67,7 @@ def run(rank, n_gpus, hps): train_sampler = DistributedBucketSampler( train_dataset, hps.train.batch_size, - [32,300,400,500,600,700,800,900,1000], + hps.data.audio_boundaries, num_replicas=n_gpus, rank=rank, shuffle=True) diff --git a/train_ms.py b/train_ms.py index 34870c62..23c8236c 100644 --- a/train_ms.py +++ b/train_ms.py @@ -67,7 +67,7 @@ def run(rank, n_gpus, hps): train_sampler = DistributedBucketSampler( train_dataset, hps.train.batch_size, - [32,300,400,500,600,700,800,900,1000], + hps.data.audio_boundaries, num_replicas=n_gpus, rank=rank, shuffle=True)