hparams/train.yaml

seed: 1993
__set_seed: !apply:torch.manual_seed [!ref <seed>]

# Dataset will be downloaded to the `data_original`
data_original: D:/voice-emo/dat/
output_folder: results/train_with_wav2vec2/1993
save_folder: !ref <output_folder>/save
train_log: !ref <output_folder>/train_log.txt

# URL for the wav2vec2 model
wav2vec2_hub: facebook/wav2vec2-base
wav2vec2_folder: !ref <save_folder>/wav2vec2_checkpoint

# Path where data manifest files will be stored
train_annotation: !ref <output_folder>/train.json
valid_annotation: !ref <output_folder>/valid.json
test_annotation: !ref <output_folder>/test.json
split_ratio: [80, 10, 10]
skip_prep: False

number_of_epochs: 5
batch_size: 4
lr: 0.0001
lr_wav2vec2: 0.00001

dataloader_options:
    batch_size: !ref <batch_size>
    shuffle: True
    num_workers: 0
    drop_last: False

encoder_dim: 768

# Number of emotions
out_n_neurons: 7  # (anger, disgust, fear, happy, neutral, sad, surprise)