| | |
|---|
| | | encoder: chunk_conformer |
|---|
| | | encoder_conf: |
|---|
| | | main_conf: |
|---|
| | | pos_wise_act_type: swish |
|---|
| | | pos_enc_dropout_rate: 0.5 |
|---|
| | | conv_mod_act_type: swish |
|---|
| | | activation_type: swish |
|---|
| | | positional_dropout_rate: 0.5 |
|---|
| | | time_reduction_factor: 2 |
|---|
| | | unified_model_training: true |
|---|
| | | default_chunk_size: 16 |
|---|
| | | jitter_range: 4 |
|---|
| | | left_chunk_size: 0 |
|---|
| | | input_conf: |
|---|
| | | block_type: conv2d |
|---|
| | | conv_size: 512 |
|---|
| | | embed_vgg_like: false |
|---|
| | | subsampling_factor: 4 |
|---|
| | | num_frame: 1 |
|---|
| | | body_conf: |
|---|
| | | - block_type: conformer |
|---|
| | | linear_size: 2048 |
|---|
| | | hidden_size: 512 |
|---|
| | | heads: 8 |
|---|
| | | linear_units: 2048 |
|---|
| | | output_size: 512 |
|---|
| | | attention_heads: 8 |
|---|
| | | dropout_rate: 0.5 |
|---|
| | | pos_wise_dropout_rate: 0.5 |
|---|
| | | att_dropout_rate: 0.5 |
|---|
| | | conv_mod_kernel_size: 15 |
|---|
| | | positional_dropout_rate: 0.5 |
|---|
| | | attention_dropout_rate: 0.5 |
|---|
| | | cnn_module_kernel: 15 |
|---|
| | | num_blocks: 12 |
|---|
| | | |
|---|
| | | # decoder related |
|---|
| | | decoder: rnn |
|---|
| | | decoder_conf: |
|---|
| | | rnnt_decoder: rnnt |
|---|
| | | rnnt_decoder_conf: |
|---|
| | | embed_size: 512 |
|---|
| | | hidden_size: 512 |
|---|
| | | embed_dropout_rate: 0.5 |
|---|
