You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
PaddleSpeech/examples/librispeech/asr5/conf/wavlmASR.yaml

138 lines
3.5 KiB

############################################
# Network Architecture #
############################################
freeze_wavlm: False
normalize_wav: True
output_norm: True
init_type: kaiming_uniform # !Warning: need to convergence
enc:
input_shape: 768
dnn_blocks: 2
dnn_neurons: 768
activation: True
normalization: True
dropout_rate: [0.15, 0]
ctc:
enc_n_units: 768
blank_id: 0
dropout_rate: 0.0
wavlm_params_path: "/home/ubuntu/Documents/Github/wavlm_paddle/wavlm-paddle-ft.pth"
task_cfg:
label_rate: 50.0
sample_rate: 16000
normalize: True
enable_padding: False
max_keep_size: None
max_sample_size: 250000
min_sample_size: 32000
dropout_input: 0.1
final_dropout: 0.0
dropout: 0.1
attention_dropout: 0.0
activation_dropout: 0.1
apply_mask: True
mask_length: 10
mask_prob: 0.5
mask_selection: static
mask_other: 0.0
no_mask_overlap: False
mask_channel_length: 10
mask_channel_prob: 0.0
mask_channel_selection: static
mask_channel_other: 0.0
no_mask_channel_overlap: False
feature_grad_mult: 0.0
layerdrop: 0.1
fp16: True
extractor_mode: layer_norm
encoder_layers: 12
encoder_embed_dim: 768
encoder_ffn_embed_dim: 3072
encoder_attention_heads: 12
activation_fn: gelu
encoder_layerdrop: 0.0
dropout_features: 0.0
final_dim: 768
untie_final_proj: True
layer_norm_first: True
conv_feature_layers: "[(512,10,5)] + [(512,3,2)] * 4 + [(512,2,2)] * 2"
conv_bias: False
logit_temp: 0.1
target_glu: False
mask_min_space: 1
mask_channel_min_space: 1
conv_pos: 128
conv_pos_groups: 16
latent_temp: [2.0, 0.5, 0.999995]
skip_masked: False
skip_nomask: True
###########################################
# Data #
###########################################
train_manifest: data/manifest.train
dev_manifest: data/manifest.dev
test_manifest: data/manifest.test-clean
###########################################
# Dataloader #
###########################################
vocab_filepath: data/lang_char/vocab.txt
unit_type: char
mean_std_filepath: ""
preprocess_config: conf/preprocess.yaml
sortagrad: 0 # Feed samples from shortest to longest ; -1: enabled for all epochs 0: disabled other: enabled for other epochs
batch_size: 8 # Different batch_size may cause large differences in results
maxlen_in: 51200000000 # if input length > maxlen-in batchsize is automatically reduced
maxlen_out: 160000
minibatches: 0 # for debug
batch_count: auto
batch_bins: 0
batch_frames_in: 0
batch_frames_out: 0
batch_frames_inout: 0
num_workers: 0
subsampling_factor: 1
num_encs: 1
dist_sampler: True
shortest_first: False
return_lens_rate: True
############################################
# Data Augmentation #
############################################
audio_augment: # for raw audio
sample_rate: 16000
speeds: [90, 100, 110]
###########################################
# Training #
###########################################
n_epoch: 10
accum_grad: 8
global_grad_clip: 5.0
model_scheduler: newbobscheduler
model_scheduler_conf:
improvement_threshold: 0.0025
annealing_factor: 0.8
patient: 0
model_optim: adam
model_optim_conf:
lr: 0.0001
weight_decay: 0.0
# I changed this
wavlm_optim: adam
wavlm_optim_conf:
lr: 0.00005
weight_decay: 0.0
wavlm_scheduler: constantlr
wavlm_scheduler_conf:
warmup_steps: 1000
lr_decay: 1.0
log_interval: 1
checkpoint:
kbest_n: 50
latest_n: 5