ddd
Add application file
b93970c
raw
history blame
No virus
1.35 kB
base_config:
- configs/singing/fs2.yaml
- usr/configs/midi/cascade/opencs/opencpop_statis.yaml
audio_sample_rate: 24000
hop_size: 128 # Hop size.
fft_size: 512 # FFT size.
win_size: 512 # FFT size.
fmin: 30
fmax: 12000
min_level_db: -120
binarization_args:
with_wav: true
with_spk_embed: false
with_align: true
raw_data_dir: 'data/raw/opencpop/segments'
processed_data_dir: 'xxx'
binarizer_cls: data_gen.singing.binarize.OpencpopBinarizer
binary_data_dir: 'data/binary/opencpop-midi-dp'
use_midi: true # for midi exp
use_gt_f0: false # for midi exp
use_gt_dur: false # for further midi exp
lambda_f0: 1.0
lambda_uv: 1.0
#lambda_energy: 0.1
lambda_ph_dur: 1.0
lambda_sent_dur: 1.0
lambda_word_dur: 1.0
predictor_grad: 0.1
pe_enable: false
pe_ckpt: ''
num_spk: 1
test_prefixes: [
'2044',
'2086',
'2092',
'2093',
'2100',
]
task_cls: usr.diffsinger_task.AuxDecoderMIDITask
#vocoder: usr.singingvocoder.highgan.HighGAN
#vocoder_ckpt: checkpoints/h_2_model/checkpoint-530000steps.pkl
vocoder: vocoders.hifigan.HifiGAN
vocoder_ckpt: checkpoints/0109_hifigan_bigpopcs_hop128
use_nsf: true
# config for experiments
max_frames: 5000
max_tokens: 40000
predictor_layers: 5
rel_pos: true
dur_predictor_layers: 5 # *
use_spk_embed: false
num_valid_plots: 10
max_updates: 160000
save_gt: true