uploaded bass model
Browse files- bass.ckpt +3 -0
- hparams.yaml +158 -0
bass.ckpt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:4f88c9437d7cdb4bb9d323c70e2ea02d76309b22d185bb41e870c0089ab58178
|
3 |
+
size 534931469
|
hparams.yaml
ADDED
@@ -0,0 +1,158 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
model:
|
2 |
+
sr: 44100
|
3 |
+
n_fft: 2048
|
4 |
+
bandsplits:
|
5 |
+
- - 500
|
6 |
+
- 50
|
7 |
+
- - 1000
|
8 |
+
- 100
|
9 |
+
- - 4000
|
10 |
+
- 500
|
11 |
+
- - 8000
|
12 |
+
- 1000
|
13 |
+
- - 16000
|
14 |
+
- 2000
|
15 |
+
bottleneck_layer: rnn
|
16 |
+
t_timesteps: 263
|
17 |
+
fc_dim: 128
|
18 |
+
rnn_dim: 256
|
19 |
+
rnn_type: LSTM
|
20 |
+
bidirectional: true
|
21 |
+
num_layers: 12
|
22 |
+
mlp_dim: 512
|
23 |
+
return_mask: false
|
24 |
+
complex_as_channel: true
|
25 |
+
is_mono: false
|
26 |
+
train_dataset:
|
27 |
+
file_dir: /home/clandschoot_kirkegaard/Music-Demixing-with-Band-Split-RNN/datasets/Bass
|
28 |
+
txt_dir: files/
|
29 |
+
txt_path: null
|
30 |
+
target: bass
|
31 |
+
is_training: true
|
32 |
+
is_mono: false
|
33 |
+
sr: 44100
|
34 |
+
preload_dataset: false
|
35 |
+
silent_prob: 0.1
|
36 |
+
mix_prob: 0.25
|
37 |
+
mix_tgt_too: false
|
38 |
+
test_dataset:
|
39 |
+
in_fp: /home/clandschoot_kirkegaard/Music-Demixing-with-Band-Split-RNN/datasets/Bass
|
40 |
+
target: bass
|
41 |
+
is_mono: false
|
42 |
+
sr: 44100
|
43 |
+
win_size: 3
|
44 |
+
hop_size: 0.5
|
45 |
+
batch_size: 4
|
46 |
+
window: null
|
47 |
+
sad:
|
48 |
+
sr: 44100
|
49 |
+
window_size_in_sec: 6
|
50 |
+
overlap_ratio: 0.5
|
51 |
+
n_chunks_per_segment: 10
|
52 |
+
eps: 1.0e-05
|
53 |
+
gamma: 0.001
|
54 |
+
threshold_max_quantile: 0.15
|
55 |
+
threshold_segment: 0.5
|
56 |
+
augmentations:
|
57 |
+
randomcrop:
|
58 |
+
_target_: data.augmentations.RandomCrop
|
59 |
+
p: 1
|
60 |
+
chunk_size_sec: 3
|
61 |
+
sr: 44100
|
62 |
+
window_stft: 2048
|
63 |
+
hop_stft: 512
|
64 |
+
gainscale:
|
65 |
+
_target_: data.augmentations.GainScale
|
66 |
+
p: 0.5
|
67 |
+
min_db: -10.0
|
68 |
+
max_db: 10.0
|
69 |
+
featurizer:
|
70 |
+
direct_transform:
|
71 |
+
_target_: torchaudio.transforms.Spectrogram
|
72 |
+
n_fft: 2048
|
73 |
+
win_length: 2048
|
74 |
+
hop_length: 512
|
75 |
+
power: null
|
76 |
+
inverse_transform:
|
77 |
+
_target_: torchaudio.transforms.InverseSpectrogram
|
78 |
+
n_fft: 2048
|
79 |
+
win_length: 2048
|
80 |
+
hop_length: 512
|
81 |
+
callbacks:
|
82 |
+
lr_monitor:
|
83 |
+
_target_: pytorch_lightning.callbacks.LearningRateMonitor
|
84 |
+
logging_interval: epoch
|
85 |
+
model_ckpt:
|
86 |
+
_target_: pytorch_lightning.callbacks.ModelCheckpoint
|
87 |
+
monitor: train/loss
|
88 |
+
mode: min
|
89 |
+
save_top_k: 5
|
90 |
+
dirpath: /home/clandschoot_kirkegaard/Music-Demixing-with-Band-Split-RNN/src/logs/bandsplitrnn/2023-04-29_14-45/weights
|
91 |
+
filename: epoch{epoch:02d}-train_loss{train/loss:.2f}
|
92 |
+
auto_insert_metric_name: false
|
93 |
+
model_ckpt_usdr:
|
94 |
+
_target_: pytorch_lightning.callbacks.ModelCheckpoint
|
95 |
+
monitor: train/usdr
|
96 |
+
mode: max
|
97 |
+
save_top_k: 5
|
98 |
+
dirpath: /home/clandschoot_kirkegaard/Music-Demixing-with-Band-Split-RNN/src/logs/bandsplitrnn/2023-04-29_14-45/weights
|
99 |
+
filename: epoch{epoch:02d}-train_usdr{train/usdr:.2f}
|
100 |
+
auto_insert_metric_name: false
|
101 |
+
ema:
|
102 |
+
_target_: utils.callbacks.EMA
|
103 |
+
decay: 0.9999
|
104 |
+
validate_original_weights: false
|
105 |
+
every_n_steps: 1
|
106 |
+
logger:
|
107 |
+
tensorboard:
|
108 |
+
_target_: pytorch_lightning.loggers.TensorBoardLogger
|
109 |
+
save_dir: /home/clandschoot_kirkegaard/Music-Demixing-with-Band-Split-RNN/src/logs/bandsplitrnn/2023-04-29_14-45/tb_logs
|
110 |
+
name: ''
|
111 |
+
version: ''
|
112 |
+
log_graph: false
|
113 |
+
default_hp_metric: false
|
114 |
+
prefix: ''
|
115 |
+
wandb:
|
116 |
+
_target_: pytorch_lightning.loggers.WandbLogger
|
117 |
+
project: MDX_BSRNN_23
|
118 |
+
name: bass
|
119 |
+
save_dir: wandb_logs
|
120 |
+
offline: false
|
121 |
+
id: null
|
122 |
+
log_model: false
|
123 |
+
prefix: ''
|
124 |
+
job_type: train
|
125 |
+
group: ''
|
126 |
+
tags: []
|
127 |
+
train_loader:
|
128 |
+
batch_size: 8
|
129 |
+
num_workers: 12
|
130 |
+
shuffle: true
|
131 |
+
drop_last: true
|
132 |
+
val_loader:
|
133 |
+
batch_size: 2
|
134 |
+
num_workers: 8
|
135 |
+
shuffle: false
|
136 |
+
drop_last: false
|
137 |
+
opt:
|
138 |
+
_target_: torch.optim.Adam
|
139 |
+
lr: 0.001
|
140 |
+
sch:
|
141 |
+
warmup_step: 10
|
142 |
+
alpha: 0.1
|
143 |
+
gamma: 0.9899494936611665
|
144 |
+
ckpt_path: logs/bandsplitrnn/2023-04-28_23-46/weights/epoch287-train_usdr8.16.ckpt
|
145 |
+
trainer:
|
146 |
+
fast_dev_run: false
|
147 |
+
min_epochs: 100
|
148 |
+
max_epochs: 500
|
149 |
+
log_every_n_steps: 10
|
150 |
+
accelerator: auto
|
151 |
+
devices: auto
|
152 |
+
gradient_clip_val: 5
|
153 |
+
precision: 32
|
154 |
+
enable_progress_bar: true
|
155 |
+
benchmark: true
|
156 |
+
deterministic: false
|
157 |
+
experiment_dirname: bandsplitrnn
|
158 |
+
wandb_api_key: d5c4447e39b2b10b95f05f907d57845ded16bc13
|