train_hf_wav2vec_full.yaml

# ################################
# Model: Wav2Vec + DNN + CTC + Softmax
# Authors:
# Gaëlle Laperrière 2023
# ################################

# ------ Paths and parameters

# Seed needs to be set at top of yaml, before objects with parameters are made.
seed: 4242
__set_seed: !apply:torch.manual_seed [!ref <seed>]
output_folder: !ref results/media_SLU_wav2vec_full/<seed>
cer_file_test: !ref <output_folder>/cer_test.txt
ctc_file_test: !ref <output_folder>/ctc_test.txt
coer_file_test: !ref <output_folder>/coer_test.txt
cver_file_test: !ref <output_folder>/cver_test.txt
save_folder: !ref <output_folder>/save
train_log: !ref <output_folder>/train_log.txt

data_folder: !PLACEHOLDER # Path of folders S0272 and E0024, to process ELRA original xml datasets.
channels_path: !PLACEHOLDER # Path of the channels.csv file downloaded via https://www.dropbox.com/sh/y7ab0lktbylz647/AADMsowYHmNYwaoL_hQt7NMha?dl=0
concepts_path: !PLACEHOLDER # Path of the concepts_full_relax.csv file downloaded via https://www.dropbox.com/sh/y7ab0lktbylz647/AADMsowYHmNYwaoL_hQt7NMha?dl=0
skip_wav: False # Skip the wav files storing if already done before.
method: full # Remove or keep specifiers in concepts. Method used by default.
task: slu # Parse SLU or ASR data.
skip_prep: False # Skip data preparation to csv because already done.
process_test2: False # Process the test2 corpus

# See https://github.com/pytorch/fairseq/blob/main/examples/wav2vec/README.md
# for Wav2Vec models and https://huggingface.co/LeBenchmark for French ones.
wav2vec_url: LeBenchmark/wav2vec2-FR-3K-large

# Data files:
csv_train: !ref <save_folder>/csv/train.csv
csv_valid: !ref <save_folder>/csv/dev.csv
csv_test: !ref <save_folder>/csv/test.csv # If the test2 was processed, you can change the file to test2.csv

# Data parameters:
# With data_parallel batch_size is split into N jobs.
# With DDP batch_size is multiplied by N jobs.
batch_size: 4
test_batch_size: 2
# We remove utterances longer than 90s in the train/dev/test sets as
# longer sentences certainly correspond to "open microphones".
avoid_if_longer_than: 90.0
avoid_if_smaller_than: 0.0
num_workers: 3
dataloader_options:
    batch_size: !ref <batch_size>
    num_workers: !ref <num_workers>
    shuffle: True
test_dataloader_options:
    batch_size: !ref <test_batch_size>
    num_workers: !ref <num_workers>

# Feature parameters:
sample_rate: 16000
feats_dim: 1024

####################### Training Parameters ####################################:
number_of_epochs: 30
lr: 1
lr_wav2vec: 0.0001
annealing_factor: 0.8
annealing_factor_wav2vec: 0.9
improvement_threshold: 0.0025
improvement_threshold_wav2vec: 0.0025
patient: 0
patient_wav2vec: 0
sorting: ascending

####################### Model Parameters #######################################
activation: !name:torch.nn.LeakyReLU
dnn_blocks: 3
dnn_neurons: 512

# Wav2Vec parameters:
freeze: False

# Decoding parameters:
blank_index: 0

# Outputs:
output_neurons: 212

# ------ Functions and classes

epoch_counter: !new:speechbrain.utils.epoch_loop.EpochCounter
    limit: !ref <number_of_epochs>

wav2vec2: !new:speechbrain.lobes.models.huggingface_transformers.wav2vec2.Wav2Vec2
    source: !ref <wav2vec_url>
    output_norm: True
    freeze: !ref <freeze>
    save_path: !ref <save_folder>/wav2vec2_checkpoint

enc: !new:speechbrain.lobes.models.VanillaNN.VanillaNN
    input_shape: [null, null, !ref <feats_dim>]
    activation: !ref <activation>
    dnn_blocks: !ref <dnn_blocks>
    dnn_neurons: !ref <dnn_neurons>

output_lin: !new:speechbrain.nnet.linear.Linear
    input_size: !ref <dnn_neurons>
    n_neurons: !ref <output_neurons>
    bias: True

log_softmax: !new:speechbrain.nnet.activations.Softmax
    apply_log: True

ctc_cost: !name:speechbrain.nnet.losses.ctc_loss
    blank_index: !ref <blank_index>

modules:
    wav2vec2: !ref <wav2vec2>
    enc: !ref <enc>
    output_lin: !ref <output_lin>

model: !new:torch.nn.ModuleList
    - [!ref <enc>, !ref <output_lin>]

model_wav2vec2: !new:torch.nn.ModuleList
    - [!ref <wav2vec2>]

opt_class: !name:torch.optim.Adadelta
    lr: !ref <lr>
    rho: 0.95
    eps: 1.e-8

opt_class_wav2vec: !name:torch.optim.Adam
    lr: !ref <lr_wav2vec>

lr_annealing: !new:speechbrain.nnet.schedulers.NewBobScheduler
    initial_value: !ref <lr>
    improvement_threshold: !ref <improvement_threshold>
    annealing_factor: !ref <annealing_factor>
    patient: !ref <patient>

lr_annealing_wav2vec: !new:speechbrain.nnet.schedulers.NewBobScheduler
    initial_value: !ref <lr_wav2vec>
    improvement_threshold: !ref <improvement_threshold_wav2vec>
    annealing_factor: !ref <annealing_factor_wav2vec>
    patient: !ref <patient_wav2vec>

label_encoder: !new:speechbrain.dataio.encoder.CTCTextEncoder

checkpointer: !new:speechbrain.utils.checkpoints.Checkpointer
    checkpoints_dir: !ref <save_folder>
    recoverables:
        model: !ref <model>
        wav2vec2: !ref <wav2vec2>
        lr_annealing: !ref <lr_annealing>
        lr_annealing_wav2vec: !ref <lr_annealing_wav2vec>
        counter: !ref <epoch_counter>
        tokenizer: !ref <label_encoder>

train_logger: !new:speechbrain.utils.train_logger.FileTrainLogger
    save_file: !ref <train_log>

ctc_computer: !name:speechbrain.utils.metric_stats.MetricStats
    metric: !name:speechbrain.nnet.losses.ctc_loss
        blank_index: !ref <blank_index>
        reduction: batch

cer_computer: !name:speechbrain.utils.metric_stats.ErrorRateStats

coer_computer: !name:speechbrain.utils.metric_stats.ErrorRateStats
    extract_concepts_values: True
    keep_values: False
    tag_in: '<'
    tag_out: '>'

cver_computer: !name:speechbrain.utils.metric_stats.ErrorRateStats
    extract_concepts_values: True
    keep_values: True
    tag_in: '<'
    tag_out: '>'