diff --git a/examples/asr/conf/carnelinet/carnelinet_384.yaml b/examples/asr/conf/carnelinet/carnelinet_384.yaml index 16369fc1a366..f0d9fadaca14 100644 --- a/examples/asr/conf/carnelinet/carnelinet_384.yaml +++ b/examples/asr/conf/carnelinet/carnelinet_384.yaml @@ -35,6 +35,8 @@ model: use_start_end_token: false max_duration: 16.7 shuffle: true + num_workers: 8 + pin_memory: true # tarred datasets is_tarred: false tarred_audio_filepaths: null @@ -50,6 +52,8 @@ model: batch_size: 32 shuffle: false use_start_end_token: false + num_workers: 8 + pin_memory: true test_ds: manifest_filepath: null @@ -57,6 +61,8 @@ model: batch_size: 32 shuffle: false use_start_end_token: false + num_workers: 8 + pin_memory: true model_defaults: repeat: 5 diff --git a/examples/asr/conf/citrinet/citrinet_1024.yaml b/examples/asr/conf/citrinet/citrinet_1024.yaml index 5c016f084ed0..79dcf5dd3e02 100644 --- a/examples/asr/conf/citrinet/citrinet_1024.yaml +++ b/examples/asr/conf/citrinet/citrinet_1024.yaml @@ -25,6 +25,8 @@ model: max_duration: 20.0 shuffle: true use_start_end_token: false + num_workers: 8 + pin_memory: true # tarred datasets is_tarred: false tarred_audio_filepaths: null @@ -39,6 +41,8 @@ model: batch_size: 32 shuffle: false use_start_end_token: false + num_workers: 8 + pin_memory: true test_ds: manifest_filepath: null @@ -46,6 +50,8 @@ model: batch_size: 32 shuffle: false use_start_end_token: false + num_workers: 8 + pin_memory: true model_defaults: repeat: 5 diff --git a/examples/asr/conf/citrinet/citrinet_384.yaml b/examples/asr/conf/citrinet/citrinet_384.yaml index 1bea46187136..ab24137b7cf5 100644 --- a/examples/asr/conf/citrinet/citrinet_384.yaml +++ b/examples/asr/conf/citrinet/citrinet_384.yaml @@ -24,6 +24,8 @@ model: max_duration: 16.7 shuffle: true use_start_end_token: false + num_workers: 8 + pin_memory: true # tarred datasets is_tarred: false tarred_audio_filepaths: null @@ -39,6 +41,8 @@ model: batch_size: 32 shuffle: false use_start_end_token: false + num_workers: 8 + pin_memory: true test_ds: manifest_filepath: null @@ -46,6 +50,8 @@ model: batch_size: 32 shuffle: false use_start_end_token: false + num_workers: 8 + pin_memory: true model_defaults: repeat: 5 diff --git a/examples/asr/conf/citrinet/config_bpe.yaml b/examples/asr/conf/citrinet/config_bpe.yaml index 07ea4c431225..2ac55618230b 100644 --- a/examples/asr/conf/citrinet/config_bpe.yaml +++ b/examples/asr/conf/citrinet/config_bpe.yaml @@ -12,6 +12,8 @@ model: trim_silence: True max_duration: 16.7 shuffle: True + num_workers: 8 + pin_memory: true # tarred datasets is_tarred: false tarred_audio_filepaths: null @@ -26,6 +28,8 @@ model: sample_rate: 16000 batch_size: 32 shuffle: False + num_workers: 8 + pin_memory: true tokenizer: dir: ??? # path to directory which contains either tokenizer.model (bpe) or vocab.txt (for wpe) diff --git a/examples/asr/conf/config.yaml b/examples/asr/conf/config.yaml index 0d283df150a4..c1ac98a0480d 100644 --- a/examples/asr/conf/config.yaml +++ b/examples/asr/conf/config.yaml @@ -15,6 +15,8 @@ model: trim_silence: True max_duration: 16.7 shuffle: True + num_workers: 8 + pin_memory: true # tarred datasets is_tarred: false tarred_audio_filepaths: null @@ -29,6 +31,8 @@ model: labels: *labels batch_size: 32 shuffle: False + num_workers: 8 + pin_memory: true preprocessor: _target_: nemo.collections.asr.modules.AudioToMelSpectrogramPreprocessor diff --git a/examples/asr/conf/contextnet_rnnt/config_rnnt.yaml b/examples/asr/conf/contextnet_rnnt/config_rnnt.yaml index 52ac211298fd..23616f75affe 100644 --- a/examples/asr/conf/contextnet_rnnt/config_rnnt.yaml +++ b/examples/asr/conf/contextnet_rnnt/config_rnnt.yaml @@ -15,6 +15,8 @@ model: max_duration: 16.7 labels: ${model.labels} shuffle: true + num_workers: 8 + pin_memory: true # tarred datasets is_tarred: false tarred_audio_filepaths: null @@ -30,6 +32,8 @@ model: batch_size: 32 shuffle: false labels: ${model.labels} + num_workers: 8 + pin_memory: true test_ds: manifest_filepath: null @@ -37,6 +41,8 @@ model: batch_size: 32 shuffle: false labels: ${model.labels} + num_workers: 8 + pin_memory: true model_defaults: repeat: 5 diff --git a/examples/asr/conf/contextnet_rnnt/config_rnnt_bpe.yaml b/examples/asr/conf/contextnet_rnnt/config_rnnt_bpe.yaml index a13b4aa3d92b..9a2581f9eda7 100644 --- a/examples/asr/conf/contextnet_rnnt/config_rnnt_bpe.yaml +++ b/examples/asr/conf/contextnet_rnnt/config_rnnt_bpe.yaml @@ -16,6 +16,8 @@ model: max_duration: 16.7 labels: [] shuffle: true + num_workers: 8 + pin_memory: true # tarred datasets is_tarred: false tarred_audio_filepaths: null @@ -30,6 +32,8 @@ model: batch_size: 32 shuffle: false labels: [] + num_workers: 8 + pin_memory: true test_ds: manifest_filepath: null @@ -37,6 +41,8 @@ model: batch_size: 32 shuffle: false labels: [] + num_workers: 8 + pin_memory: true model_defaults: repeat: 5 diff --git a/examples/asr/conf/jasper/jasper_10x5dr.yaml b/examples/asr/conf/jasper/jasper_10x5dr.yaml index 85c18a66005c..42ead2e6ae4c 100644 --- a/examples/asr/conf/jasper/jasper_10x5dr.yaml +++ b/examples/asr/conf/jasper/jasper_10x5dr.yaml @@ -13,6 +13,8 @@ model: trim_silence: True max_duration: 16.7 shuffle: True + num_workers: 8 + pin_memory: true # tarred datasets is_tarred: false tarred_audio_filepaths: null @@ -28,6 +30,8 @@ model: labels: *labels batch_size: 32 shuffle: False + num_workers: 8 + pin_memory: true preprocessor: _target_: nemo.collections.asr.modules.AudioToMelSpectrogramPreprocessor diff --git a/examples/asr/conf/marblenet/marblenet_3x2x64.yaml b/examples/asr/conf/marblenet/marblenet_3x2x64.yaml index 319feae0fcb4..22c10972697e 100644 --- a/examples/asr/conf/marblenet/marblenet_3x2x64.yaml +++ b/examples/asr/conf/marblenet/marblenet_3x2x64.yaml @@ -19,6 +19,8 @@ model: tarred_audio_filepaths: null tarred_shard_strategy: "scatter" shuffle_n: 2048 + num_workers: 8 + pin_memory: true # bucketing params bucketing_strategy: "synced_randomized" bucketing_batch_size: null @@ -38,6 +40,8 @@ model: labels: ${model.labels} batch_size: 128 shuffle: False + num_workers: 8 + pin_memory: true val_loss_idx: 0 test_ds: @@ -46,6 +50,8 @@ model: labels: ${model.labels} batch_size: 128 shuffle: False + num_workers: 8 + pin_memory: true test_loss_idx: 0 preprocessor: diff --git a/examples/asr/conf/matchboxnet/matchboxnet_3x1x64_v1.yaml b/examples/asr/conf/matchboxnet/matchboxnet_3x1x64_v1.yaml index 73d2aa5b8c52..687d68ea98d0 100644 --- a/examples/asr/conf/matchboxnet/matchboxnet_3x1x64_v1.yaml +++ b/examples/asr/conf/matchboxnet/matchboxnet_3x1x64_v1.yaml @@ -21,6 +21,8 @@ model: labels: ${model.labels} batch_size: 128 shuffle: True + num_workers: 8 + pin_memory: true # tarred datasets is_tarred: false tarred_audio_filepaths: null @@ -44,6 +46,8 @@ model: labels: ${model.labels} batch_size: 128 shuffle: False + num_workers: 8 + pin_memory: true val_loss_idx: 0 test_ds: @@ -52,6 +56,8 @@ model: labels: ${model.labels} batch_size: 128 shuffle: False + num_workers: 8 + pin_memory: true test_loss_idx: 0 preprocessor: diff --git a/examples/asr/conf/matchboxnet/matchboxnet_3x1x64_v2.yaml b/examples/asr/conf/matchboxnet/matchboxnet_3x1x64_v2.yaml index 5af861fbb29f..54a28b4c2060 100644 --- a/examples/asr/conf/matchboxnet/matchboxnet_3x1x64_v2.yaml +++ b/examples/asr/conf/matchboxnet/matchboxnet_3x1x64_v2.yaml @@ -21,6 +21,8 @@ model: labels: ${model.labels} batch_size: 128 shuffle: True + num_workers: 8 + pin_memory: true # tarred datasets is_tarred: false tarred_audio_filepaths: null @@ -44,6 +46,8 @@ model: labels: ${model.labels} batch_size: 128 shuffle: False + num_workers: 8 + pin_memory: true val_loss_idx: 0 test_ds: @@ -52,6 +56,8 @@ model: labels: ${model.labels} batch_size: 128 shuffle: False + num_workers: 8 + pin_memory: true test_loss_idx: 0 preprocessor: diff --git a/examples/asr/conf/quartznet/quartznet_15x5.yaml b/examples/asr/conf/quartznet/quartznet_15x5.yaml index 80136a2de585..3c0699a1bdc0 100644 --- a/examples/asr/conf/quartznet/quartznet_15x5.yaml +++ b/examples/asr/conf/quartznet/quartznet_15x5.yaml @@ -16,6 +16,8 @@ model: trim_silence: True max_duration: 16.7 shuffle: True + num_workers: 8 + pin_memory: true # tarred datasets is_tarred: false tarred_audio_filepaths: null @@ -30,6 +32,8 @@ model: labels: *labels batch_size: 32 shuffle: False + num_workers: 8 + pin_memory: true test_ds: manifest_filepath: null @@ -37,6 +41,8 @@ model: labels: *labels batch_size: 32 shuffle: False + num_workers: 8 + pin_memory: true preprocessor: _target_: nemo.collections.asr.modules.AudioToMelSpectrogramPreprocessor diff --git a/examples/asr/conf/ssl/citrinet/citrinet_ssl_1024.yaml b/examples/asr/conf/ssl/citrinet/citrinet_ssl_1024.yaml index 1e8289842561..7ab5902f245a 100644 --- a/examples/asr/conf/ssl/citrinet/citrinet_ssl_1024.yaml +++ b/examples/asr/conf/ssl/citrinet/citrinet_ssl_1024.yaml @@ -28,6 +28,8 @@ model: tarred_audio_filepaths: null shuffle_n: 2048 use_start_end_token: false + num_workers: 8 + pin_memory: true # bucketing params bucketing_strategy: "synced_randomized" bucketing_batch_size: null @@ -38,6 +40,8 @@ model: batch_size: 32 shuffle: false use_start_end_token: false + num_workers: 8 + pin_memory: true max_duration: 35.0 min_duration: 8.0 diff --git a/examples/asr/conf/ssl/citrinet/citrinet_ssl_ci.yaml b/examples/asr/conf/ssl/citrinet/citrinet_ssl_ci.yaml index 4c2a55ac1113..4ea2a98347a8 100644 --- a/examples/asr/conf/ssl/citrinet/citrinet_ssl_ci.yaml +++ b/examples/asr/conf/ssl/citrinet/citrinet_ssl_ci.yaml @@ -16,6 +16,8 @@ model: is_tarred: false tarred_audio_filepaths: null use_start_end_token: false + num_workers: 8 + pin_memory: true # bucketing params bucketing_strategy: "synced_randomized" bucketing_batch_size: null diff --git a/examples/asr/conf/wav2vec/wav2vecCTC.yaml b/examples/asr/conf/wav2vec/wav2vecCTC.yaml index 89d97aa2e5e1..11c9576e6f6d 100644 --- a/examples/asr/conf/wav2vec/wav2vecCTC.yaml +++ b/examples/asr/conf/wav2vec/wav2vecCTC.yaml @@ -19,6 +19,8 @@ model: is_tarred: false tarred_audio_filepaths: null use_start_end_token: false + num_workers: 8 + pin_memory: true validation_ds: manifest_filepath: ??? @@ -26,6 +28,8 @@ model: batch_size: ?? shuffle: false use_start_end_token: false + num_workers: 8 + pin_memory: true test_ds: manifest_filepath: null @@ -33,6 +37,8 @@ model: batch_size: null shuffle: false use_start_end_token: false + num_workers: 8 + pin_memory: true tokenizer: dir: ??? # path to directory which contains either tokenizer.model (bpe) or vocab.txt (for wpe) diff --git a/examples/asr/conf/wav2vec/wav2vecCTC_large.yaml b/examples/asr/conf/wav2vec/wav2vecCTC_large.yaml index 911c466aa137..0ca0914acecc 100644 --- a/examples/asr/conf/wav2vec/wav2vecCTC_large.yaml +++ b/examples/asr/conf/wav2vec/wav2vecCTC_large.yaml @@ -18,6 +18,8 @@ model: is_tarred: false tarred_audio_filepaths: null use_start_end_token: false + num_workers: 8 + pin_memory: true validation_ds: manifest_filepath: ??? @@ -25,6 +27,8 @@ model: batch_size: 4 shuffle: false use_start_end_token: false + num_workers: 8 + pin_memory: true test_ds: manifest_filepath: null @@ -32,6 +36,8 @@ model: batch_size: null shuffle: false use_start_end_token: false + num_workers: 8 + pin_memory: true tokenizer: dir: ??? # path to directory which contains either tokenizer.model (bpe) or vocab.txt (for wpe) diff --git a/examples/asr/conf/wav2vec/wav2vec_pretrain.yaml b/examples/asr/conf/wav2vec/wav2vec_pretrain.yaml index 836294fbeef2..0aaad93be4c1 100644 --- a/examples/asr/conf/wav2vec/wav2vec_pretrain.yaml +++ b/examples/asr/conf/wav2vec/wav2vec_pretrain.yaml @@ -24,6 +24,8 @@ model: is_tarred: false tarred_audio_filepaths: null use_start_end_token: false + num_workers: 8 + pin_memory: true validation_ds: manifest_filepath: ??? @@ -31,6 +33,8 @@ model: batch_size: ??? shuffle: false use_start_end_token: false + num_workers: 8 + pin_memory: true preprocessor: _target_: nemo.collections.asr.modules.wav2vec_modules.ConvFeatureEncoder diff --git a/examples/asr/conf/wav2vec/wav2vec_pretrain_large.yaml b/examples/asr/conf/wav2vec/wav2vec_pretrain_large.yaml index c1d74cf4d29d..b69dade0d98d 100644 --- a/examples/asr/conf/wav2vec/wav2vec_pretrain_large.yaml +++ b/examples/asr/conf/wav2vec/wav2vec_pretrain_large.yaml @@ -23,6 +23,8 @@ model: is_tarred: false tarred_audio_filepaths: null use_start_end_token: false + num_workers: 8 + pin_memory: true validation_ds: manifest_filepath: ??? @@ -30,6 +32,8 @@ model: batch_size: ??? shuffle: false use_start_end_token: false + num_workers: 8 + pin_memory: true preprocessor: _target_: nemo.collections.asr.modules.wav2vec_modules.ConvFeatureEncoder diff --git a/examples/asr/experimental/k2/conf/citrinet/citrinet_mmi_1024.yaml b/examples/asr/experimental/k2/conf/citrinet/citrinet_mmi_1024.yaml index 2784ef6fb68d..c8fa1a52caab 100644 --- a/examples/asr/experimental/k2/conf/citrinet/citrinet_mmi_1024.yaml +++ b/examples/asr/experimental/k2/conf/citrinet/citrinet_mmi_1024.yaml @@ -25,6 +25,8 @@ model: max_duration: 20.0 shuffle: true use_start_end_token: false + num_workers: 8 + pin_memory: true # tarred datasets is_tarred: false tarred_audio_filepaths: null @@ -39,6 +41,8 @@ model: batch_size: 32 shuffle: false use_start_end_token: false + num_workers: 8 + pin_memory: true test_ds: manifest_filepath: null @@ -46,6 +50,8 @@ model: batch_size: 32 shuffle: false use_start_end_token: false + num_workers: 8 + pin_memory: true model_defaults: repeat: 5 diff --git a/examples/asr/experimental/wav2vec/configs/wav2vecCTC.yaml b/examples/asr/experimental/wav2vec/configs/wav2vecCTC.yaml index 09a4ddc4a51c..d31b554d2dc3 100644 --- a/examples/asr/experimental/wav2vec/configs/wav2vecCTC.yaml +++ b/examples/asr/experimental/wav2vec/configs/wav2vecCTC.yaml @@ -19,6 +19,8 @@ model: is_tarred: false tarred_audio_filepaths: null use_start_end_token: false + num_workers: 8 + pin_memory: true validation_ds: manifest_filepath: ??? @@ -26,6 +28,8 @@ model: batch_size: ?? shuffle: false use_start_end_token: false + num_workers: 8 + pin_memory: true test_ds: manifest_filepath: null @@ -33,6 +37,8 @@ model: batch_size: null shuffle: false use_start_end_token: false + num_workers: 8 + pin_memory: true tokenizer: dir: ??? # path to directory which contains either tokenizer.model (bpe) or vocab.txt (for wpe) diff --git a/examples/asr/experimental/wav2vec/configs/wav2vecCTC_large.yaml b/examples/asr/experimental/wav2vec/configs/wav2vecCTC_large.yaml index f6121bc369f8..93d751f8d1ce 100644 --- a/examples/asr/experimental/wav2vec/configs/wav2vecCTC_large.yaml +++ b/examples/asr/experimental/wav2vec/configs/wav2vecCTC_large.yaml @@ -18,6 +18,8 @@ model: is_tarred: false tarred_audio_filepaths: null use_start_end_token: false + num_workers: 8 + pin_memory: true validation_ds: manifest_filepath: ??? @@ -25,6 +27,8 @@ model: batch_size: 4 shuffle: false use_start_end_token: false + num_workers: 8 + pin_memory: true test_ds: manifest_filepath: null @@ -32,6 +36,8 @@ model: batch_size: null shuffle: false use_start_end_token: false + num_workers: 8 + pin_memory: true tokenizer: dir: ??? # path to directory which contains either tokenizer.model (bpe) or vocab.txt (for wpe) diff --git a/examples/asr/experimental/wav2vec/configs/wav2vec_pretrain.yaml b/examples/asr/experimental/wav2vec/configs/wav2vec_pretrain.yaml index e929e9d38a6b..8e4bc6c18f8f 100644 --- a/examples/asr/experimental/wav2vec/configs/wav2vec_pretrain.yaml +++ b/examples/asr/experimental/wav2vec/configs/wav2vec_pretrain.yaml @@ -24,6 +24,8 @@ model: is_tarred: false tarred_audio_filepaths: null use_start_end_token: false + num_workers: 8 + pin_memory: true validation_ds: manifest_filepath: ??? @@ -31,6 +33,8 @@ model: batch_size: ??? shuffle: false use_start_end_token: false + num_workers: 8 + pin_memory: true preprocessor: _target_: nemo.collections.asr.modules.wav2vec_modules.ConvFeatureEncoder diff --git a/examples/asr/experimental/wav2vec/configs/wav2vec_pretrain_large.yaml b/examples/asr/experimental/wav2vec/configs/wav2vec_pretrain_large.yaml index a569cccfdd3d..fa211e7b7b44 100644 --- a/examples/asr/experimental/wav2vec/configs/wav2vec_pretrain_large.yaml +++ b/examples/asr/experimental/wav2vec/configs/wav2vec_pretrain_large.yaml @@ -23,6 +23,8 @@ model: is_tarred: false tarred_audio_filepaths: null use_start_end_token: false + num_workers: 8 + pin_memory: true validation_ds: manifest_filepath: ??? @@ -30,6 +32,8 @@ model: batch_size: ??? shuffle: false use_start_end_token: false + num_workers: 8 + pin_memory: true preprocessor: _target_: nemo.collections.asr.modules.wav2vec_modules.ConvFeatureEncoder