audio_to_text/audiocaps_cntrstv_cnn14rnn_trm/config.yaml

model:
    encoder:
        type: Cnn14RnnEncoder
        args:
            sample_rate: 32000
            pretrained: ./audio_to_text/pretrained_feature_extractors/contrastive_pretrain_cnn14_bertm.pth
            freeze_cnn: True
            freeze_cnn_bn: True
            bidirectional: True
            dropout: 0.5
            hidden_size: 256
            num_layers: 3
    decoder:
        type: TransformerDecoder
        args:
            attn_emb_dim: 512
            dropout: 0.2
            emb_dim: 256
            fc_emb_dim: 512
            nlayers: 2
    type: TransformerModel
    args: {}