hyperparams.yaml

# Generated 2023-10-18 from:
# /home/wakeb/Abdulrahman-tts/speechbrain/speechbrain/recipes/VoxCeleb/SpeakerRec/hparams/train_x_vectors.yaml
# yamllint disable
# ################################
# Model: Speaker identification with ECAPA
# Authors: Hwidong Na & Mirco Ravanelli
# ################################

# Basic parameters
seed: 651
__set_seed: !apply:torch.manual_seed [651]
output_folder: /media/wakeb/T7 Touch/speechbrain651
save_folder: /media/wakeb/T7 Touch/speechbrain651/save
train_log: /media/wakeb/T7 Touch/speechbrain651/train_log.txt

# Data files
data_folder: /media/wakeb/T7 Touch/data_qasr/  # e.g. /path/to/Voxceleb
train_annotation: /media/wakeb/T7 Touch/data_qasr/train_reworked.csv
valid_annotation: /media/wakeb/T7 Touch/data_qasr/dev_reworked.csv

# Folder to extract data augmentation files
rir_folder: /media/wakeb/T7 Touch/data_qasr/ # Change it if needed

# Use the following links for the official voxceleb splits:
# VoxCeleb1 (cleaned): https://www.robots.ox.ac.uk/~vgg/data/voxceleb/meta/veri_test2.txt
# VoxCeleb1-H (cleaned): https://www.robots.ox.ac.uk/~vgg/data/voxceleb/meta/list_test_hard2.txt
# VoxCeleb1-E (cleaned): https://www.robots.ox.ac.uk/~vgg/data/voxceleb/meta/list_test_all2.txt.
# VoxCeleb1-E and VoxCeleb1-H lists are drawn from the VoxCeleb1 training set.
# Therefore you cannot use any files in VoxCeleb1 for training if you are using these lists for testing.
verification_file: /media/wakeb/T7 Touch/data_qasr/testing.txt

split_ratio: [90, 10]
skip_prep: true
ckpt_interval_minutes: 15 # save checkpoint every N min

# Training parameters
number_of_epochs: 30
batch_size: 140
lr: 0.001
lr_final: 0.000001

sample_rate: 16000
sentence_len: 3 # seconds
shuffle: false
random_chunk: true

# Feature parameters
n_mels: 24
left_frames: 0
right_frames: 0
deltas: false

# Number of speakers
out_n_neurons: 2917
emb_dim: 512

dataloader_options:
  batch_size: 140
  shuffle: false
  num_workers: 0

# Functions
compute_features: &id006 !new:speechbrain.lobes.features.Fbank
  n_mels: 24
  left_frames: 0
  right_frames: 0
  deltas: false

embedding_model: &id007 !new:speechbrain.lobes.models.Xvector.Xvector
  in_channels: 24
  activation: !name:torch.nn.LeakyReLU
  tdnn_blocks: 5
  tdnn_channels: [512, 512, 512, 512, 1500]
  tdnn_kernel_sizes: [5, 3, 3, 1, 1]
  tdnn_dilations: [1, 2, 3, 1, 1]
  lin_neurons: 512

classifier: &id008 !new:speechbrain.lobes.models.Xvector.Classifier
  input_shape: [null, null, 512]
  activation: !name:torch.nn.LeakyReLU
  lin_blocks: 1
  lin_neurons: 512
  out_neurons: 2917

epoch_counter: &id010 !new:speechbrain.utils.epoch_loop.EpochCounter
  limit: 30


augment_wavedrop: &id001 !new:speechbrain.lobes.augment.TimeDomainSpecAugment
  sample_rate: 16000
  speeds: [100]

augment_speed: &id002 !new:speechbrain.lobes.augment.TimeDomainSpecAugment
  sample_rate: 16000
  speeds: [95, 100, 105]

add_rev: &id003 !new:speechbrain.lobes.augment.EnvCorrupt
  openrir_folder: /media/wakeb/T7 Touch/data_qasr/
  openrir_max_noise_len: 3.0    # seconds
  reverb_prob: 1.0
  noise_prob: 0.0
  noise_snr_low: 0
  noise_snr_high: 15
  rir_scale_factor: 1.0

add_noise: &id004 !new:speechbrain.lobes.augment.EnvCorrupt
  openrir_folder: /media/wakeb/T7 Touch/data_qasr/
  openrir_max_noise_len: 3.0    # seconds
  reverb_prob: 0.0
  noise_prob: 1.0
  noise_snr_low: 0
  noise_snr_high: 15
  rir_scale_factor: 1.0

add_rev_noise: &id005 !new:speechbrain.lobes.augment.EnvCorrupt
  openrir_folder: /media/wakeb/T7 Touch/data_qasr/
  openrir_max_noise_len: 3.0    # seconds
  reverb_prob: 1.0
  noise_prob: 1.0
  noise_snr_low: 0
  noise_snr_high: 15
  rir_scale_factor: 1.0


# Definition of the augmentation pipeline.
# If concat_augment = False, the augmentation techniques are applied
# in sequence. If concat_augment = True, all the augmented signals
# are concatenated in a single big batch.
augment_pipeline: [*id001, *id002, *id003, *id004, *id005]
concat_augment: true

mean_var_norm: &id009 !new:speechbrain.processing.features.InputNormalization

# Cost + optimization
  norm_type: sentence
  std_norm: false

modules:
  compute_features: *id006
  augment_wavedrop: *id001
  augment_speed: *id002
  add_rev: *id003
  add_noise: *id004
  add_rev_noise: *id005
  embedding_model: *id007
  classifier: *id008
  mean_var_norm: *id009
compute_cost: !name:speechbrain.nnet.losses.nll_loss
# compute_error: !name:speechbrain.nnet.losses.classification_error

opt_class: !name:torch.optim.Adam
  lr: 0.001
  weight_decay: 0.000002

lr_annealing: !new:speechbrain.nnet.schedulers.LinearScheduler
  initial_value: 0.001
  final_value: 0.000001
  epoch_count: 30

# Logging + checkpoints
train_logger: !new:speechbrain.utils.train_logger.FileTrainLogger
  save_file: /media/wakeb/T7 Touch/speechbrain651/train_log.txt

error_stats: !name:speechbrain.utils.metric_stats.MetricStats
  metric: !name:speechbrain.nnet.losses.classification_error
    reduction: batch

checkpointer: !new:speechbrain.utils.checkpoints.Checkpointer
  checkpoints_dir: /media/wakeb/T7 Touch/speechbrain651/save
  recoverables:
    embedding_model: *id007
    classifier: *id008
    normalizer: *id009
    counter: *id010