subatomicseer
commited on
Commit
•
27de6d7
1
Parent(s):
6b163f1
Upload 2 files
Browse files- config.yaml +146 -0
- pytorch_model.bin +3 -0
config.yaml
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# pytorch_lightning==1.8.6
|
2 |
+
seed_everything: 4444
|
3 |
+
trainer:
|
4 |
+
logger:
|
5 |
+
class_path: pytorch_lightning.loggers.TensorBoardLogger
|
6 |
+
init_args:
|
7 |
+
save_dir: logs/
|
8 |
+
name: lightning_logs
|
9 |
+
version: null
|
10 |
+
log_graph: false
|
11 |
+
default_hp_metric: true
|
12 |
+
prefix: ''
|
13 |
+
sub_dir: null
|
14 |
+
logdir: null
|
15 |
+
comment: ''
|
16 |
+
purge_step: null
|
17 |
+
max_queue: 10
|
18 |
+
flush_secs: 120
|
19 |
+
filename_suffix: ''
|
20 |
+
write_to_disk: true
|
21 |
+
comet_config:
|
22 |
+
disabled: true
|
23 |
+
enable_checkpointing: true
|
24 |
+
callbacks:
|
25 |
+
- class_path: pytorch_lightning.callbacks.LearningRateMonitor
|
26 |
+
init_args:
|
27 |
+
logging_interval: null
|
28 |
+
log_momentum: false
|
29 |
+
- class_path: pytorch_lightning.callbacks.ModelSummary
|
30 |
+
init_args:
|
31 |
+
max_depth: 2
|
32 |
+
- class_path: pytorch_lightning.callbacks.ModelCheckpoint
|
33 |
+
init_args:
|
34 |
+
dirpath: null
|
35 |
+
filename: vocos_checkpoint_{epoch}_{step}_{val_loss:.4f}
|
36 |
+
monitor: val_loss
|
37 |
+
verbose: false
|
38 |
+
save_last: true
|
39 |
+
save_top_k: 3
|
40 |
+
save_weights_only: false
|
41 |
+
mode: min
|
42 |
+
auto_insert_metric_name: true
|
43 |
+
every_n_train_steps: null
|
44 |
+
train_time_interval: null
|
45 |
+
every_n_epochs: null
|
46 |
+
save_on_train_epoch_end: null
|
47 |
+
- class_path: vocos.helpers.GradNormCallback
|
48 |
+
default_root_dir: null
|
49 |
+
gradient_clip_val: null
|
50 |
+
gradient_clip_algorithm: null
|
51 |
+
num_nodes: 1
|
52 |
+
num_processes: null
|
53 |
+
devices:
|
54 |
+
- 0
|
55 |
+
- 1
|
56 |
+
- 2
|
57 |
+
gpus: null
|
58 |
+
auto_select_gpus: false
|
59 |
+
tpu_cores: null
|
60 |
+
ipus: null
|
61 |
+
enable_progress_bar: true
|
62 |
+
overfit_batches: 0.0
|
63 |
+
track_grad_norm: -1
|
64 |
+
check_val_every_n_epoch: 1
|
65 |
+
fast_dev_run: false
|
66 |
+
accumulate_grad_batches: null
|
67 |
+
max_epochs: null
|
68 |
+
min_epochs: null
|
69 |
+
max_steps: 2000000
|
70 |
+
min_steps: null
|
71 |
+
max_time: null
|
72 |
+
limit_train_batches: null
|
73 |
+
limit_val_batches: 100
|
74 |
+
limit_test_batches: null
|
75 |
+
limit_predict_batches: null
|
76 |
+
val_check_interval: null
|
77 |
+
log_every_n_steps: 100
|
78 |
+
accelerator: gpu
|
79 |
+
strategy: ddp
|
80 |
+
sync_batchnorm: false
|
81 |
+
precision: 32
|
82 |
+
enable_model_summary: true
|
83 |
+
num_sanity_val_steps: 2
|
84 |
+
resume_from_checkpoint: null
|
85 |
+
profiler: null
|
86 |
+
benchmark: null
|
87 |
+
deterministic: null
|
88 |
+
reload_dataloaders_every_n_epochs: 0
|
89 |
+
auto_lr_find: false
|
90 |
+
replace_sampler_ddp: true
|
91 |
+
detect_anomaly: false
|
92 |
+
auto_scale_batch_size: false
|
93 |
+
plugins: null
|
94 |
+
amp_backend: native
|
95 |
+
amp_level: null
|
96 |
+
move_metrics_to_cpu: false
|
97 |
+
multiple_trainloader_mode: max_size_cycle
|
98 |
+
inference_mode: true
|
99 |
+
data:
|
100 |
+
class_path: vocos.dataset.VocosDataModule
|
101 |
+
init_args:
|
102 |
+
train_params:
|
103 |
+
filelist_path: /data/umeiro0/datasets/librispeech/LibriSpeech/filelist.train
|
104 |
+
base_dir: /data/umeiro0/datasets/librispeech/LibriSpeech
|
105 |
+
sampling_rate: 16000
|
106 |
+
num_samples: 16000
|
107 |
+
batch_size: 32
|
108 |
+
num_workers: 8
|
109 |
+
val_params:
|
110 |
+
filelist_path: /data/umeiro0/datasets/librispeech/LibriSpeech/filelist.val
|
111 |
+
base_dir: /data/umeiro0/datasets/librispeech/LibriSpeech
|
112 |
+
sampling_rate: 16000
|
113 |
+
num_samples: 16000
|
114 |
+
batch_size: 16
|
115 |
+
num_workers: 8
|
116 |
+
model:
|
117 |
+
class_path: vocos.experiment.VocosDACExp
|
118 |
+
init_args:
|
119 |
+
feature_extractor:
|
120 |
+
class_path: vocos.feature_extractors.DACFeatures
|
121 |
+
backbone:
|
122 |
+
class_path: vocos.models.VocosBackbone
|
123 |
+
init_args:
|
124 |
+
input_channels: 1024
|
125 |
+
dim: 384
|
126 |
+
intermediate_dim: 1152
|
127 |
+
num_layers: 8
|
128 |
+
layer_scale_init_value: null
|
129 |
+
adanorm_num_embeddings: null
|
130 |
+
head:
|
131 |
+
class_path: vocos.heads.ISTFTHead
|
132 |
+
init_args:
|
133 |
+
dim: 384
|
134 |
+
n_fft: 1280
|
135 |
+
hop_length: 320
|
136 |
+
padding: same
|
137 |
+
sample_rate: 16000
|
138 |
+
initial_learning_rate: 0.0005
|
139 |
+
num_warmup_steps: 0
|
140 |
+
mel_loss_coeff: 45.0
|
141 |
+
mrd_loss_coeff: 1.0
|
142 |
+
pretrain_mel_steps: 0
|
143 |
+
decay_mel_coeff: false
|
144 |
+
evaluate_utmos: true
|
145 |
+
evaluate_pesq: true
|
146 |
+
evaluate_periodicty: true
|
pytorch_model.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b89183291b503f246227dd145947839400e7f509c12a70764967f2117bd24580
|
3 |
+
size 508782458
|