s3prl-vc-vcc2020 / TEF2 /config.yml
unilight's picture
Upload with huggingface_hub
d6b61d7
additional_config: null
allow_cache: true
batch_size: 6
config: conf/taco2_ar.yaml
dev_scp: data/TEF2_dev/wav.scp
distributed: false
eval_interval_steps: 1000
fft_size: 1024
fmax: 7600
fmin: 80
global_gain_scale: 1.0
grad_norm: 1.0
hop_size: 256
init_checkpoint: ''
log_interval_steps: 100
main_loss_type: L1Loss
model_params:
ar: true
encoder_type: taco2
hidden_dim: 1024
lstmp_dropout_rate: 0.2
lstmp_layernorm: false
lstmp_layers: 2
lstmp_proj_dim: 256
prenet_dim: 256
prenet_dropout_rate: 0.5
prenet_layers: 2
model_type: Taco2_AR
num_mels: 80
num_save_intermediate_results: 4
num_workers: 2
optimizer_params:
lr: 0.0001
optimizer_type: AdamW
outdir: exp/TEF2_ppg_sxliu_taco2_ar
pin_memory: true
rank: 0
resume: ''
sampling_rate: 24000
save_interval_steps: 1000
scheduler: linear_schedule_with_warmup
scheduler_params:
num_warmup_steps: 4000
train_max_steps: 10000
train_scp: data/TEF2_train/wav.scp
trg_stats: exp/TEF2_ppg_sxliu_taco2_ar/stats.h5
trim_frame_size: 2048
trim_hop_size: 512
trim_silence: false
trim_threshold_in_db: 60
upstream: ppg_sxliu
verbose: 1
version: 0.1.5
vocoder:
checkpoint: ./downloads/pwg_task1/checkpoint-400000steps.pkl
config: ./downloads/pwg_task1/config.yml
stats: ./downloads/pwg_task1/stats.h5
win_length: null
window: hann