|
|
@@ -0,0 +1,104 @@
|
|
|
+# network architecture
|
|
|
+# encoder related
|
|
|
+encoder: branchformer
|
|
|
+encoder_conf:
|
|
|
+ output_size: 256
|
|
|
+ use_attn: true
|
|
|
+ attention_heads: 4
|
|
|
+ attention_layer_type: rel_selfattn
|
|
|
+ pos_enc_layer_type: rel_pos
|
|
|
+ rel_pos_type: latest
|
|
|
+ use_cgmlp: true
|
|
|
+ cgmlp_linear_units: 2048
|
|
|
+ cgmlp_conv_kernel: 31
|
|
|
+ use_linear_after_conv: false
|
|
|
+ gate_activation: identity
|
|
|
+ merge_method: concat
|
|
|
+ cgmlp_weight: 0.5 # used only if merge_method is "fixed_ave"
|
|
|
+ attn_branch_drop_rate: 0.0 # used only if merge_method is "learned_ave"
|
|
|
+ num_blocks: 24
|
|
|
+ dropout_rate: 0.1
|
|
|
+ positional_dropout_rate: 0.1
|
|
|
+ attention_dropout_rate: 0.1
|
|
|
+ input_layer: conv2d
|
|
|
+ stochastic_depth_rate: 0.0
|
|
|
+
|
|
|
+# decoder related
|
|
|
+decoder: transformer
|
|
|
+decoder_conf:
|
|
|
+ attention_heads: 4
|
|
|
+ linear_units: 2048
|
|
|
+ num_blocks: 6
|
|
|
+ dropout_rate: 0.1
|
|
|
+ positional_dropout_rate: 0.1
|
|
|
+ self_attention_dropout_rate: 0.
|
|
|
+ src_attention_dropout_rate: 0.
|
|
|
+
|
|
|
+# frontend related
|
|
|
+frontend: wav_frontend
|
|
|
+frontend_conf:
|
|
|
+ fs: 16000
|
|
|
+ window: hamming
|
|
|
+ n_mels: 80
|
|
|
+ frame_length: 25
|
|
|
+ frame_shift: 10
|
|
|
+ lfr_m: 1
|
|
|
+ lfr_n: 1
|
|
|
+
|
|
|
+# hybrid CTC/attention
|
|
|
+model_conf:
|
|
|
+ ctc_weight: 0.3
|
|
|
+ lsm_weight: 0.1 # label smoothing option
|
|
|
+ length_normalized_loss: false
|
|
|
+
|
|
|
+# optimization related
|
|
|
+accum_grad: 1
|
|
|
+grad_clip: 5
|
|
|
+max_epoch: 180
|
|
|
+val_scheduler_criterion:
|
|
|
+ - valid
|
|
|
+ - acc
|
|
|
+best_model_criterion:
|
|
|
+- - valid
|
|
|
+ - acc
|
|
|
+ - max
|
|
|
+keep_nbest_models: 10
|
|
|
+
|
|
|
+optim: adam
|
|
|
+optim_conf:
|
|
|
+ lr: 0.001
|
|
|
+ weight_decay: 0.000001
|
|
|
+scheduler: warmuplr
|
|
|
+scheduler_conf:
|
|
|
+ warmup_steps: 35000
|
|
|
+
|
|
|
+specaug: specaug
|
|
|
+specaug_conf:
|
|
|
+ apply_time_warp: true
|
|
|
+ time_warp_window: 5
|
|
|
+ time_warp_mode: bicubic
|
|
|
+ apply_freq_mask: true
|
|
|
+ freq_mask_width_range:
|
|
|
+ - 0
|
|
|
+ - 27
|
|
|
+ num_freq_mask: 2
|
|
|
+ apply_time_mask: true
|
|
|
+ time_mask_width_ratio_range:
|
|
|
+ - 0.
|
|
|
+ - 0.05
|
|
|
+ num_time_mask: 10
|
|
|
+
|
|
|
+dataset_conf:
|
|
|
+ data_names: speech,text
|
|
|
+ data_types: sound,text
|
|
|
+ shuffle: True
|
|
|
+ shuffle_conf:
|
|
|
+ shuffle_size: 2048
|
|
|
+ sort_size: 500
|
|
|
+ batch_conf:
|
|
|
+ batch_type: token
|
|
|
+ batch_size: 10000
|
|
|
+ num_workers: 8
|
|
|
+
|
|
|
+log_interval: 50
|
|
|
+normalize: None
|
嘉渊