model_class: enc_dec_rnnt_bpe sample_rate: 16000 log_prediction: true model_defaults: enc_hidden: 768 pred_hidden: 320 join_hidden: 320 preprocessor: _target_: __main__.AudioToMelSpectrogramPreprocessor sample_rate: 16000 n_fft: 400 n_window_size: 400 window_size: null n_window_stride: 160 window_stride: null features: 64 dither: 0.0 preemph: null log: true log_zero_guard_type: clamp normalize: null pad_to: 0 mel_norm: null window: hann log_zero_guard_value: 1e-9 tokenizer: dir: tokenizer_all_sets/ type: bpe validation_ds: shuffle: False manifest_filepath: null encoder: _target_: nemo.collections.asr.modules.ConformerEncoder feat_in: 64 feat_out: -1 n_layers: 16 d_model: 768 subsampling: striding subsampling_factor: 4 subsampling_conv_channels: 768 ff_expansion_factor: 4 self_attention_model: rel_pos pos_emb_max_len: 5000 n_heads: 16 xscaling: false untie_biases: true conv_kernel_size: 31 dropout: 0.1 dropout_emb: 0.1 dropout_att: 0.1 decoder: _target_: nemo.collections.asr.modules.RNNTDecoder normalization_mode: null random_state_sampling: false blank_as_pad: true vocab_size: 512 prednet: pred_hidden: 320 pred_rnn_layers: 1 t_max: null dropout: 0.0 joint: _target_: nemo.collections.asr.modules.RNNTJoint log_softmax: null fuse_loss_wer: false fused_batch_size: 1 jointnet: joint_hidden: 320 activation: relu dropout: 0.0 encoder_hidden: 768 optim: name: adamw lr: 5.0e-05 betas: - 0.9 - 0.98 weight_decay: 0.01 sched: name: CosineAnnealing warmup_steps: 10000 warmup_ratio: null min_lr: 1.0e-07 nemo_version: 1.12.0 decoding: strategy: greedy_batch preserve_alignments: false greedy: max_symbols: 3 beam: beam_size: 5 score_norm: true loss: loss_name: default mwer: false rnnt_reduction: mean_batch wer_coef: false subtract_mean: true warprnnt_numba_kwargs: fastemit_lambda: 0.0 clamp: -1.0 rnnt_weight: 0.1 unique_hyp: true