Upload lsc_THXC_ppg_sxliu/config.yml with huggingface_hub
Browse files- lsc_THXC_ppg_sxliu/config.yml +172 -0
lsc_THXC_ppg_sxliu/config.yml
ADDED
@@ -0,0 +1,172 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
additional_config: conf/vtn.tts_pt.v1.ppg_sxliu.yaml
|
2 |
+
allow_cache: true
|
3 |
+
batch_size: 16
|
4 |
+
cleaner: tacotron
|
5 |
+
collater_type: ARVCCollater
|
6 |
+
config: exp/TXHC_bdl_1032_debug/original_config.yml
|
7 |
+
criterions:
|
8 |
+
Seq2SeqLoss:
|
9 |
+
bce_pos_weight: 10.0
|
10 |
+
dev_dp_input_dir: null
|
11 |
+
dev_dumpdir: dump/dev/ppg_sxliu_norm
|
12 |
+
dev_duration_dir: null
|
13 |
+
dev_text: data/dev/text
|
14 |
+
distributed: false
|
15 |
+
eval_interval_steps: 500
|
16 |
+
feat_type: ppg_sxliu
|
17 |
+
fft_size: 1024
|
18 |
+
fmax: 7600
|
19 |
+
fmin: 80
|
20 |
+
format: hdf5
|
21 |
+
freeze-mods: null
|
22 |
+
g2p: g2p_en
|
23 |
+
global_gain_scale: 1.0
|
24 |
+
grad_norm: 1.0
|
25 |
+
hop_size: 256
|
26 |
+
inference:
|
27 |
+
maxlenratio: 6.0
|
28 |
+
minlenratio: 0.0
|
29 |
+
threshold: 0.5
|
30 |
+
init-mods:
|
31 |
+
- encoder
|
32 |
+
- decoder
|
33 |
+
- feat_out
|
34 |
+
- prob_out
|
35 |
+
- postnet
|
36 |
+
init_checkpoint: exp/TXHC_bdl_1032_debug/original_checkpoint-50000steps.pkl
|
37 |
+
log_interval_steps: 10
|
38 |
+
model_params:
|
39 |
+
adim: 384
|
40 |
+
aheads: 4
|
41 |
+
decoder_concat_after: false
|
42 |
+
decoder_normalize_before: false
|
43 |
+
decoder_reduction_factor: 2
|
44 |
+
dlayers: 6
|
45 |
+
dprenet_layers: 2
|
46 |
+
dprenet_units: 256
|
47 |
+
dunits: 1536
|
48 |
+
elayers: 6
|
49 |
+
encoder_concat_after: false
|
50 |
+
encoder_normalize_before: true
|
51 |
+
eunits: 1536
|
52 |
+
idim: 144
|
53 |
+
odim: 144
|
54 |
+
postnet_chans: 256
|
55 |
+
postnet_filts: 5
|
56 |
+
postnet_layers: 5
|
57 |
+
use_batch_norm: true
|
58 |
+
model_type: VTN
|
59 |
+
non_linguistic_symbols: none
|
60 |
+
num_mels: 80
|
61 |
+
num_save_intermediate_results: 4
|
62 |
+
num_workers: 2
|
63 |
+
optimizer_params:
|
64 |
+
lr: 8.0e-05
|
65 |
+
optimizer_type: Adam
|
66 |
+
outdir: exp/TXHC_bdl_1032_debug
|
67 |
+
pin_memory: true
|
68 |
+
pretrain: ''
|
69 |
+
rank: 0
|
70 |
+
resume: ''
|
71 |
+
sampling_rate: 16000
|
72 |
+
save_interval_steps: 1000
|
73 |
+
scheduler: warmuplr
|
74 |
+
scheduler_params:
|
75 |
+
warmup_steps: 4000
|
76 |
+
seq2seq_loss_params:
|
77 |
+
bce_pos_weight: 10.0
|
78 |
+
src_dev_dumpdir: dump/TXHC_dev/ppg_sxliu/norm_ljspeech
|
79 |
+
src_feat_type: ppg_sxliu
|
80 |
+
src_train_dumpdir: dump/TXHC_train_1032/ppg_sxliu/norm_ljspeech
|
81 |
+
stats:
|
82 |
+
mean: !!python/object/apply:numpy.core.multiarray._reconstruct
|
83 |
+
args:
|
84 |
+
- &id001 !!python/name:numpy.ndarray ''
|
85 |
+
- !!python/tuple
|
86 |
+
- 0
|
87 |
+
- !!binary |
|
88 |
+
Yg==
|
89 |
+
state: !!python/tuple
|
90 |
+
- 1
|
91 |
+
- !!python/tuple
|
92 |
+
- 144
|
93 |
+
- &id002 !!python/object/apply:numpy.dtype
|
94 |
+
args:
|
95 |
+
- f4
|
96 |
+
- false
|
97 |
+
- true
|
98 |
+
state: !!python/tuple
|
99 |
+
- 3
|
100 |
+
- <
|
101 |
+
- null
|
102 |
+
- null
|
103 |
+
- null
|
104 |
+
- -1
|
105 |
+
- -1
|
106 |
+
- 0
|
107 |
+
- false
|
108 |
+
- !!binary |
|
109 |
+
6RzTPXVNQb7seW+/uWClPQCf5DwPGXQ9G1WePo3UFz6KNp4+FW29PKUMYb5cODu+UbHTPWSOQL53
|
110 |
+
uc48guEYPn4qgT6KxCO/ER4aPgElyT4BsV6+jpz4valJYj54Z4q+di9MvkSt3z2KlVi/lI+kPVRf
|
111 |
+
DD+rzu8+CC1svkHG2ruaX9I9tog+vniP5j2YLva8L589vhSeSL/l4PE9s+6VPJZ2szsyuQi+1Opi
|
112 |
+
PU7OzbxdUNc/RtTqPfLVnL5ASu0+fp6mPuvjLryKgqU9p4hVu2In9T2D1ri+vAz5PHVzzz3ac+G9
|
113 |
+
Ecu9PZx+Hj6Mrx49kWyAPfqRNj6s4WW+BjmJPlp7+L0eN1g+UzhKPMuUkr7yhl29DwY8viWowbyS
|
114 |
+
iUe9igiJPX+YgT5CNy++JExQvirydT1kAdA9V3TjvcHclz2aklg+QtSYPahhl75+c9U9yzjjvHAm
|
115 |
+
973nkJm9KkJEPvg9dL2KZYE+GmnxvGzyk75R9c8+KuMGvqXd6r3sxZg9Nv0jPmTSS76EHfe90mGq
|
116 |
+
Pd2alb0BxbW+HNUcPxoIwD16mQ++PkZCvhh8jT5I6Kg/HT1pPUMkWj4dO4I9dePAu3qRqb1s/Zo+
|
117 |
+
G+Mrvf4WIz+DW7o+yHIePsTqRry9aGk+HVjvPRnqc7xE9S495SlKvOMueD9qdKG+kabJvaV17z3B
|
118 |
+
Q5g8VVjYuvwZbz22gwi+B98XPSBKIz8+dVc+Gq+FPcAYlr301Vw9L0y2vQj5pL36o9Q+sgqHPs++
|
119 |
+
3z1xMCQ/
|
120 |
+
scale: !!python/object/apply:numpy.core.multiarray._reconstruct
|
121 |
+
args:
|
122 |
+
- *id001
|
123 |
+
- !!python/tuple
|
124 |
+
- 0
|
125 |
+
- !!binary |
|
126 |
+
Yg==
|
127 |
+
state: !!python/tuple
|
128 |
+
- 1
|
129 |
+
- !!python/tuple
|
130 |
+
- 144
|
131 |
+
- *id002
|
132 |
+
- false
|
133 |
+
- !!binary |
|
134 |
+
OanzPqWL3T4RC9M+kg0VP0g2ED8rY/E+UDTvPrfl2T7xOOA+C97MPp688D4/LvE+r3rbPgGi5z47
|
135 |
+
MNI+VOHEPgbm2D5gJs0+higRP8sS5z7TvQY/y8rXPnlO1j42lvg++sr0PocB5D5JMtU+2mj0PkBu
|
136 |
+
HD8joiU/1ibePq5jyD4ReO4+CE7YPgbv6j79B7s+euHxPobUxz7c8vc+DhicP9qf6z4SVQA/UpHx
|
137 |
+
Pu6tzD7MChM/1wj3Po76xz4WyEY/wtvdPkRM6T7xdfg+YGXfPnrHAj984tQ+EFTqPqSQCz+O3RI/
|
138 |
+
Osq/PsgW1j4aDts+Obr9Ppw45z7nNOU+J9DyPt4D+T468Ao/jzEaP9Fq5j7Gb+U+NALCPrPy3j7U
|
139 |
+
BA8/Nl73PiCq0T5qr+Y+FaQVP8+d5j5Ave0+jx7oPlKI6j4q5+U+f//IPqW6DD+UWSE/3WPJPix+
|
140 |
+
2D4+kgU/TnYWP7GX1z55iOo+qHP1PgIh3D6zgD4/MHEJP7zp1z5Aqtw+k1X2Pvynyj5PJ88+lhjX
|
141 |
+
PrSi2j5F9fk+du3ZPggsyD6yNNE+lmrDPujPNz8Kscc+mx/0Ppaw/T5UoOk+pjHpPgpgxz5flAI/
|
142 |
+
MnjSPo190j7X2t8+gmvpPlgnFD9q0CY/u8oQP4ER+D536dc+9LfJPnkQAD/rKMI+i4YIP7xyBz+B
|
143 |
+
nf0+VcTKPqbVAT/ck+I+9uTePhdvAz/UGuo+ahfTPt0/9D7shcQ++VnSPphNzz4Ewvk+DlAoP+91
|
144 |
+
9j4qV/4+
|
145 |
+
token_list: exp/phn_tacotron_transformer_tts.v1.ppg_sxliu/tokens.txt
|
146 |
+
token_type: phn
|
147 |
+
train_dp_input_dir: null
|
148 |
+
train_dumpdir: dump/train_no_dev/ppg_sxliu_norm
|
149 |
+
train_duration_dir: null
|
150 |
+
train_max_steps: 50000
|
151 |
+
train_text: data/train_no_dev/text
|
152 |
+
trainer_type: ARVCTrainer
|
153 |
+
trg_dev_dumpdir: dump/bdl_dev/ppg_sxliu/norm_ljspeech
|
154 |
+
trg_feat_type: ppg_sxliu
|
155 |
+
trg_stats: exp/TXHC_bdl_1032_debug/stats.h5
|
156 |
+
trg_train_dumpdir: dump/bdl_train_1032/ppg_sxliu/norm_ljspeech
|
157 |
+
trim_frame_size: 2048
|
158 |
+
trim_hop_size: 512
|
159 |
+
trim_silence: false
|
160 |
+
trim_threshold_in_db: 60
|
161 |
+
verbose: 1
|
162 |
+
version: 0.1.0
|
163 |
+
vocoder:
|
164 |
+
# checkpoint: /data/group1/z44476r/Experiments/s3prl-vc/egs/l2-arctic/a2o_vc/exp/TXHC_ppg_sxliu_taco2_ar/checkpoint-10000steps.pkl
|
165 |
+
# config: /data/group1/z44476r/Experiments/s3prl-vc/egs/l2-arctic/a2o_vc/exp/TXHC_ppg_sxliu_taco2_ar/config.yml
|
166 |
+
# stats: /data/group1/z44476r/Experiments/s3prl-vc/egs/l2-arctic/a2o_vc/exp/TXHC_ppg_sxliu_taco2_ar/stats.h5
|
167 |
+
checkpoint: /data/group1/z44476r/Experiments/seq2seq-vc/egs/l2-arctic/lsc/downloads/ppg_sxliu_decoder_plus_pwg_THXC/checkpoint-10000steps.pkl
|
168 |
+
config: /data/group1/z44476r/Experiments/seq2seq-vc/egs/l2-arctic/lsc/downloads/ppg_sxliu_decoder_plus_pwg_THXC/config.yml
|
169 |
+
stats: /data/group1/z44476r/Experiments/seq2seq-vc/egs/l2-arctic/lsc/downloads/ppg_sxliu_decoder_plus_pwg_THXC/stats.h5
|
170 |
+
vocoder_type: s3prl_vc
|
171 |
+
win_length: null
|
172 |
+
window: hann
|