diffvox / presets /fx_config.yaml
yoyolicoris's picture
manually copy part of the diffvox source code
d737ecd
epochs: 2000
data_dir: null
log_dir: null
lufs: -18
sr: 44100
chunk_duration: 12
chunk_overlap: 5
device: cuda
batch_size: 35
dataset: medley_vocal
regularise_delay: true
model:
_target_: torch.nn.Sequential
_args_:
- _target_: modules.fx.Peak
sr: 44100
freq: 800
min_freq: 33
max_freq: 5400
- _target_: modules.fx.Peak
sr: 44100
freq: 4000
min_freq: 200
max_freq: 17500
- _target_: modules.fx.LowShelf
sr: 44100
freq: 115
min_freq: 30
max_freq: 200
- _target_: modules.fx.HighShelf
sr: 44100
freq: 6000
min_freq: 750
max_freq: 8300
- _target_: modules.fx.LowPass
sr: 44100
freq: 17500
min_freq: 200
max_freq: 18000
- _target_: modules.fx.HighPass
sr: 44100
freq: 200
min_freq: 16
max_freq: 5300
- _target_: modules.fx.CompressorExpander
sr: 44100
cmp_ratio: 2.0
exp_ratio: 0.5
at_ms: 50.0
rt_ms: 50.0
avg_coef: 0.3
cmp_th: -18.0
exp_th: -48.0
make_up: 0.0
lookahead: true
max_lookahead: 15
- _target_: modules.fx.SendFXsAndSum
_args_:
- _target_: modules.fx.SurrogateDelay
sr: 44100
delay: 400
dropout: 0
straight_through: true
recursive_eq: true
ir_duration: 4
eq:
_target_: modules.fx.LowPass
sr: 44100
freq: 8000
min_freq: 200
max_freq: 16000
min_Q: 0.5
max_Q: 2
- _target_: modules.fx.FDN
sr: 44100
delays:
- 997
- 1153
- 1327
- 1559
- 1801
- 2099
num_decay_freq: 49
delay_independent_decay: true
ir_duration: 12
eq:
_target_: torch.nn.Sequential
_args_:
- _target_: modules.fx.Peak
sr: 44100
freq: 800
min_freq: 200
max_freq: 2500
min_Q: 0.1
max_Q: 3
- _target_: modules.fx.Peak
sr: 44100
freq: 4000
min_freq: 600
max_freq: 7000
min_Q: 0.1
max_Q: 3
- _target_: modules.fx.LowShelf
sr: 44100
freq: 115
min_freq: 30
max_freq: 450
- _target_: modules.fx.HighShelf
sr: 44100
freq: 8000
min_freq: 1500
max_freq: 16000
cross_send: true
pan_direct: true
optimiser:
_target_: torch.optim.Adam
lr: 0.01
mss:
fft_sizes:
- 128
- 512
- 2048
hop_sizes:
- 32
- 128
- 512
mldr:
s_taus:
- 50
- 100
l_taus:
- 1000
- 2000
loss_fn:
_target_: loss.SumLosses
weights:
- 1.0
- 0.5
- 0.5
- 0.25
loss_fns:
- _target_: auraloss.freq.MultiResolutionSTFTLoss
fft_sizes:
- 128
- 512
- 2048
hop_sizes:
- 32
- 128
- 512
win_lengths:
- 128
- 512
- 2048
sample_rate: 44100
perceptual_weighting: true
- _target_: auraloss.freq.SumAndDifferenceSTFTLoss
fft_sizes:
- 128
- 512
- 2048
hop_sizes:
- 32
- 128
- 512
win_lengths:
- 128
- 512
- 2048
sample_rate: 44100
perceptual_weighting: true
- _target_: loss.ldr.MLDRLoss
sr: 44100
s_taus:
- 50
- 100
l_taus:
- 1000
- 2000
- _target_: loss.ldr.MLDRLoss
sr: 44100
mid_side: true
s_taus:
- 50
- 100
l_taus:
- 1000
- 2000