Introduce code from Rust monorepo. This is a mess

tts/NeMo/examples/asr/conf/matchboxnet/matchboxnet_3x1x64_v1.yaml

@@ -0,0 +1,193 @@
+name: &name "MatchboxNet-3x1x64-v1"
+
+model:
+  sample_rate: 16000
+  timesteps: 128
+  repeat: 1
+  dropout: 0.0
+  kernel_size_factor: 1.0
+
+  labels_full: ['bed', 'bird', 'cat', 'dog', 'down', 'eight', 'five', 'four', 'go', 'happy', 'house', 'left', 'marvin',
+           'nine', 'no', 'off', 'on', 'one', 'right', 'seven', 'sheila', 'six', 'stop', 'three', 'tree', 'two', 'up',
+           'wow', 'yes', 'zero']
+
+  labels_subset: ["yes", "no", "up", "down", "left", "right", "on", "off", "stop", "go", "unknown", "silence"]
+
+  labels: ${model.labels_full}
+
+  train_ds:
+    manifest_filepath: ???
+    sample_rate: ${model.sample_rate}
+    labels: ${model.labels}
+    batch_size: 128
+    shuffle: True
+    is_tarred: False
+    tarred_audio_filepaths: null
+    tarred_shard_strategy: "scatter"
+    augmentor:
+      shift:
+        prob: 1.0
+        min_shift_ms: -5.0
+        max_shift_ms: 5.0
+      white_noise:
+        prob: 1.0
+        min_level: -90
+        max_level: -46
+
+  validation_ds:
+    manifest_filepath: ???
+    sample_rate: ${model.sample_rate}
+    labels: ${model.labels}
+    batch_size: 128
+    shuffle: False
+    val_loss_idx: 0
+
+  test_ds:
+    manifest_filepath: null
+    sample_rate: ${model.sample_rate}
+    labels: ${model.labels}
+    batch_size: 128
+    shuffle: False
+    test_loss_idx: 0
+
+  preprocessor:
+    _target_: nemo.collections.asr.modules.AudioToMFCCPreprocessor
+    window_size: 0.025
+    window_stride: 0.01
+    window: "hann"
+    n_mels: &n_mels 64
+    n_mfcc: *n_mels
+    n_fft: 512
+
+  spec_augment:
+    _target_: nemo.collections.asr.modules.SpectrogramAugmentation
+    freq_masks: 2
+    time_masks: 2
+    freq_width: 15
+    time_width: 25
+    rect_masks: 5
+    rect_time: 25
+    rect_freq: 15
+
+  crop_or_pad_augment:
+    _target_: nemo.collections.asr.modules.CropOrPadSpectrogramAugmentation
+    audio_length: ${model.timesteps}
+
+  encoder:
+    _target_: nemo.collections.asr.modules.ConvASREncoder
+    feat_in: *n_mels
+    activation: relu
+    conv_mask: true
+
+    jasper:
+      - filters: 128
+        repeat: 1
+        kernel: [11]
+        stride: [1]
+        dilation: [1]
+        dropout: ${model.dropout}
+        residual: false
+        separable: true
+        kernel_size_factor: ${model.kernel_size_factor}
+
+      - filters: 64
+        repeat: ${model.repeat}
+        kernel: [13]
+        stride: [1]
+        dilation: [1]
+        dropout: ${model.dropout}
+        residual: true
+        separable: true
+        kernel_size_factor: ${model.kernel_size_factor}
+
+      - filters: 64
+        repeat: ${model.repeat}
+        kernel: [15]
+        stride: [1]
+        dilation: [1]
+        dropout: ${model.dropout}
+        residual: true
+        separable: true
+        kernel_size_factor: ${model.kernel_size_factor}
+
+      - filters: 64
+        repeat: ${model.repeat}
+        kernel: [17]
+        stride: [1]
+        dilation: [1]
+        dropout: ${model.dropout}
+        residual: true
+        separable: true
+        kernel_size_factor: ${model.kernel_size_factor}
+
+      - filters: 128
+        repeat: 1
+        kernel: [29]
+        stride: [1]
+        dilation: [2]
+        dropout: ${model.dropout}
+        residual: false
+        separable: true
+        kernel_size_factor: ${model.kernel_size_factor}
+
+      - filters: &enc_final_filters 128
+        repeat: 1
+        kernel: [1]
+        stride: [1]
+        dilation: [1]
+        dropout: ${model.dropout}
+        residual: false
+
+  decoder:
+    _target_: nemo.collections.asr.modules.ConvASRDecoderClassification
+    feat_in: *enc_final_filters
+    return_logits: true
+    pooling_type: 'avg'
+
+  optim:
+    name: novograd
+    # _target_: nemo.core.optim.optimizers.Novograd
+    lr: 0.05
+    # optimizer arguments
+    betas: [0.95, 0.5]
+    weight_decay: 0.001
+
+    # scheduler setup
+    sched:
+      name: PolynomialHoldDecayAnnealing
+
+      # Scheduler params
+      power: 2.0
+      warmup_ratio: 0.05
+      hold_ratio: 0.45
+      min_lr: 0.001
+      last_epoch: -1
+
+trainer:
+  gpus: 0 # number of gpus
+  max_epochs: 200
+  max_steps: null # computed at runtime if not set
+  num_nodes: 1
+  accelerator: ddp
+  accumulate_grad_batches: 1
+  checkpoint_callback: False  # Provided by exp_manager
+  logger: False  # Provided by exp_manager
+  log_every_n_steps: 1  # Interval of logging.
+  val_check_interval: 1.0  # Set to 0.25 to check 4 times per epoch, or an int for number of iterations
+
+exp_manager:
+  exp_dir: null
+  name: *name
+  create_tensorboard_logger: True
+  create_checkpoint_callback: True
+  create_wandb_logger: False
+  wandb_logger_kwargs:
+    name: null
+    project: null
+
+hydra:
+  run:
+    dir: .
+  job_logging:
+    root:
+      handlers: null

tts/NeMo/examples/asr/conf/matchboxnet/matchboxnet_3x1x64_v2.yaml

@@ -0,0 +1,193 @@
+name: &name "MatchboxNet-3x1x64-v2"
+
+model:
+  sample_rate: 16000
+  timesteps: 128
+  repeat: 1
+  dropout: 0.0
+  kernel_size_factor: 1.0
+
+  labels_full: ['visual', 'wow', 'learn', 'backward', 'dog', 'two', 'left', 'happy', 'nine', 'go', 'up', 'bed', 'stop',
+           'one', 'zero', 'tree', 'seven', 'on', 'four', 'bird', 'right', 'eight', 'no', 'six', 'forward', 'house',
+           'marvin', 'sheila', 'five', 'off', 'three', 'down', 'cat', 'follow', 'yes']
+
+  labels_subset: ["yes", "no", "up", "down", "left", "right", "on", "off", "stop", "go", "unknown", "silence"]
+
+  labels: ${model.labels_full}
+
+  train_ds:
+    manifest_filepath: ???
+    sample_rate: ${model.sample_rate}
+    labels: ${model.labels}
+    batch_size: 128
+    shuffle: True
+    is_tarred: False
+    tarred_audio_filepaths: null
+    tarred_shard_strategy: "scatter"
+    augmentor:
+      shift:
+        prob: 1.0
+        min_shift_ms: -5.0
+        max_shift_ms: 5.0
+      white_noise:
+        prob: 1.0
+        min_level: -90
+        max_level: -46
+
+  validation_ds:
+    manifest_filepath: ???
+    sample_rate: ${model.sample_rate}
+    labels: ${model.labels}
+    batch_size: 128
+    shuffle: False
+    val_loss_idx: 0
+
+  test_ds:
+    manifest_filepath: null
+    sample_rate: ${model.sample_rate}
+    labels: ${model.labels}
+    batch_size: 128
+    shuffle: False
+    test_loss_idx: 0
+
+  preprocessor:
+    _target_: nemo.collections.asr.modules.AudioToMFCCPreprocessor
+    window_size: 0.025
+    window_stride: 0.01
+    window: "hann"
+    n_mels: &n_mels 64
+    n_mfcc: *n_mels
+    n_fft: 512
+
+  spec_augment:
+    _target_: nemo.collections.asr.modules.SpectrogramAugmentation
+    freq_masks: 2
+    time_masks: 2
+    freq_width: 15
+    time_width: 25
+    rect_masks: 5
+    rect_time: 25
+    rect_freq: 15
+
+  crop_or_pad_augment:
+    _target_: nemo.collections.asr.modules.CropOrPadSpectrogramAugmentation
+    audio_length: ${model.timesteps}
+
+  encoder:
+    _target_: nemo.collections.asr.modules.ConvASREncoder
+    feat_in: *n_mels
+    activation: relu
+    conv_mask: true
+
+    jasper:
+      - filters: 128
+        repeat: 1
+        kernel: [11]
+        stride: [1]
+        dilation: [1]
+        dropout: ${model.dropout}
+        residual: false
+        separable: true
+        kernel_size_factor: ${model.kernel_size_factor}
+
+      - filters: 64
+        repeat: ${model.repeat}
+        kernel: [13]
+        stride: [1]
+        dilation: [1]
+        dropout: ${model.dropout}
+        residual: true
+        separable: true
+        kernel_size_factor: ${model.kernel_size_factor}
+
+      - filters: 64
+        repeat: ${model.repeat}
+        kernel: [15]
+        stride: [1]
+        dilation: [1]
+        dropout: ${model.dropout}
+        residual: true
+        separable: true
+        kernel_size_factor: ${model.kernel_size_factor}
+
+      - filters: 64
+        repeat: ${model.repeat}
+        kernel: [17]
+        stride: [1]
+        dilation: [1]
+        dropout: ${model.dropout}
+        residual: true
+        separable: true
+        kernel_size_factor: ${model.kernel_size_factor}
+
+      - filters: 128
+        repeat: 1
+        kernel: [29]
+        stride: [1]
+        dilation: [2]
+        dropout: ${model.dropout}
+        residual: false
+        separable: true
+        kernel_size_factor: ${model.kernel_size_factor}
+
+      - filters: &enc_final_filters 128
+        repeat: 1
+        kernel: [1]
+        stride: [1]
+        dilation: [1]
+        dropout: ${model.dropout}
+        residual: false
+
+  decoder:
+    _target_: nemo.collections.asr.modules.ConvASRDecoderClassification
+    feat_in: *enc_final_filters
+    return_logits: true
+    pooling_type: 'avg'
+
+  optim:
+    name: novograd
+    # _target_: nemo.core.optim.optimizers.Novograd
+    lr: 0.05
+    # optimizer arguments
+    betas: [0.95, 0.5]
+    weight_decay: 0.001
+
+    # scheduler setup
+    sched:
+      name: PolynomialHoldDecayAnnealing
+
+      # Scheduler params
+      power: 2.0
+      warmup_ratio: 0.05
+      hold_ratio: 0.45
+      min_lr: 0.001
+      last_epoch: -1
+
+trainer:
+  gpus: 0 # number of gpus
+  max_epochs: 200
+  max_steps: null # computed at runtime if not set
+  num_nodes: 1
+  accelerator: ddp
+  accumulate_grad_batches: 1
+  checkpoint_callback: False  # Provided by exp_manager
+  logger: False  # Provided by exp_manager
+  log_every_n_steps: 1  # Interval of logging.
+  val_check_interval: 1.0  # Set to 0.25 to check 4 times per epoch, or an int for number of iterations
+
+exp_manager:
+  exp_dir: null
+  name: *name
+  create_tensorboard_logger: True
+  create_checkpoint_callback: True
+  create_wandb_logger: False
+  wandb_logger_kwargs:
+    name: null
+    project: null
+
+hydra:
+  run:
+    dir: .
+  job_logging:
+    root:
+      handlers: null