forked from speechbrain/speechbrain
-
Notifications
You must be signed in to change notification settings - Fork 0
/
sepformer-whamr.yaml
192 lines (159 loc) · 5.92 KB
/
sepformer-whamr.yaml
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
# ################################
# Model: SepFormer for source separation
# https://arxiv.org/abs/2010.13154
#
# Dataset : WHAMR!
# ################################
# Basic parameters
# Seed needs to be set at top of yaml, before objects with parameters are made
#
seed: 1234
__set_seed: !apply:torch.manual_seed [!ref <seed>]
# Data params
# the data folder for the wham dataset
# data_folder needs to follow the format: /yourpath/whamr.
# make sure to use the name whamr at your top folder for the dataset!
data_folder: /yourpath/whamr
# the path for wsj0/si_tr_s/ folder -- only needed if dynamic mixing is used
# e.g. /yourpath/wsj0-processed/si_tr_s/
# you need to convert the original wsj0 to 8k
# you can do this conversion with the script ../meta/preprocess_dynamic_mixing.py
base_folder_dm: /yourpath/wsj0-processed/si_tr_s/
experiment_name: sepformer-whamr
output_folder: !ref results/<experiment_name>/<seed>
train_log: !ref <output_folder>/train_log.txt
save_folder: !ref <output_folder>/save
# the file names should start with whamr instead of whamorg
train_data: !ref <save_folder>/whamr_tr.csv
valid_data: !ref <save_folder>/whamr_cv.csv
test_data: !ref <save_folder>/whamr_tt.csv
skip_prep: False
# Experiment params
precision: fp16 # bf16, fp16 or fp32 # Set it to True for mixed precision
num_spks: 2 # set to 3 for wsj0-3mix
save_audio: False # Save estimated sources on disk
sample_rate: 8000
####################### Training Parameters ####################################
N_epochs: 200
batch_size: 1
lr: 0.00015
clip_grad_norm: 5
loss_upper_lim: 999999 # this is the upper limit for an acceptable loss
# if True, the training sequences are cut to a specified length
limit_training_signal_len: False
# this is the length of sequences if we choose to limit
# the signal length of training sequences
training_signal_len: 32000000
# Set it to True to dynamically create mixtures at training time
dynamic_mixing: False
# Parameters for data augmentation
# rir_path variable points to the directory of the room impulse responses
# e.g. /miniscratch/subakany/rir_wavs
# If the path does not exist, it is created automatically.
rir_path: !PLACEHOLDER
use_wavedrop: False
use_speedperturb: True
use_rand_shift: False
min_shift: -8000
max_shift: 8000
# Speed perturbation
speed_changes: [95, 100, 105] # List of speed changes for time-stretching
speed_perturb: !new:speechbrain.augment.time_domain.SpeedPerturb
orig_freq: !ref <sample_rate>
speeds: !ref <speed_changes>
# Frequency drop: randomly drops a number of frequency bands to zero.
drop_freq_low: 0 # Min frequency band dropout probability
drop_freq_high: 1 # Max frequency band dropout probability
drop_freq_count_low: 1 # Min number of frequency bands to drop
drop_freq_count_high: 3 # Max number of frequency bands to drop
drop_freq_width: 0.05 # Width of frequency bands to drop
drop_freq: !new:speechbrain.augment.time_domain.DropFreq
drop_freq_low: !ref <drop_freq_low>
drop_freq_high: !ref <drop_freq_high>
drop_freq_count_low: !ref <drop_freq_count_low>
drop_freq_count_high: !ref <drop_freq_count_high>
drop_freq_width: !ref <drop_freq_width>
# Time drop: randomly drops a number of temporal chunks.
drop_chunk_count_low: 1 # Min number of audio chunks to drop
drop_chunk_count_high: 5 # Max number of audio chunks to drop
drop_chunk_length_low: 1000 # Min length of audio chunks to drop
drop_chunk_length_high: 2000 # Max length of audio chunks to drop
drop_chunk: !new:speechbrain.augment.time_domain.DropChunk
drop_length_low: !ref <drop_chunk_length_low>
drop_length_high: !ref <drop_chunk_length_high>
drop_count_low: !ref <drop_chunk_count_low>
drop_count_high: !ref <drop_chunk_count_high>
# loss thresholding -- this thresholds the training loss
threshold_byloss: True
threshold: -30
# Encoder parameters
N_encoder_out: 256
out_channels: 256
kernel_size: 16
kernel_stride: 8
# Dataloader options
dataloader_opts:
batch_size: !ref <batch_size>
num_workers: 3
# Specifying the network
Encoder: !new:speechbrain.lobes.models.dual_path.Encoder
kernel_size: !ref <kernel_size>
out_channels: !ref <N_encoder_out>
SBtfintra: !new:speechbrain.lobes.models.dual_path.SBTransformerBlock
num_layers: 8
d_model: !ref <out_channels>
nhead: 8
d_ffn: 1024
dropout: 0
use_positional_encoding: True
norm_before: True
SBtfinter: !new:speechbrain.lobes.models.dual_path.SBTransformerBlock
num_layers: 8
d_model: !ref <out_channels>
nhead: 8
d_ffn: 1024
dropout: 0
use_positional_encoding: True
norm_before: True
MaskNet: !new:speechbrain.lobes.models.dual_path.Dual_Path_Model
num_spks: !ref <num_spks>
in_channels: !ref <N_encoder_out>
out_channels: !ref <out_channels>
num_layers: 2
K: 250
intra_model: !ref <SBtfintra>
inter_model: !ref <SBtfinter>
norm: ln
linear_layer_after_inter_intra: False
skip_around_intra: True
Decoder: !new:speechbrain.lobes.models.dual_path.Decoder
in_channels: !ref <N_encoder_out>
out_channels: 1
kernel_size: !ref <kernel_size>
stride: !ref <kernel_stride>
bias: False
optimizer: !name:torch.optim.Adam
lr: !ref <lr>
weight_decay: 0
loss: !name:speechbrain.nnet.losses.get_si_snr_with_pitwrapper
lr_scheduler: !new:speechbrain.nnet.schedulers.ReduceLROnPlateau
factor: 0.5
patience: 2
dont_halve_until_epoch: 85
epoch_counter: !new:speechbrain.utils.epoch_loop.EpochCounter
limit: !ref <N_epochs>
modules:
encoder: !ref <Encoder>
decoder: !ref <Decoder>
masknet: !ref <MaskNet>
save_all_checkpoints: True
checkpointer: !new:speechbrain.utils.checkpoints.Checkpointer
checkpoints_dir: !ref <save_folder>
recoverables:
encoder: !ref <Encoder>
decoder: !ref <Decoder>
masknet: !ref <MaskNet>
counter: !ref <epoch_counter>
lr_scheduler: !ref <lr_scheduler>
train_logger: !new:speechbrain.utils.train_logger.FileTrainLogger
save_file: !ref <train_log>