NeMo
Xinsheng-Wang's picture
Upload folder using huggingface_hub
83dc502 verified
accumulate_grad_batches: 1
amp: false
audio_num_mel_bins: 80
audio_sample_rate: 24000
base_config:
- ./base.yaml
- ./rosvot.yaml
binarization_args:
min_sil_duration: 0.1
shuffle: false
test_range:
- 0
- 100
train_range:
- 200
- -1
trim_eos_bos: false
valid_range:
- 100
- 200
with_align: true
with_f0: true
with_f0cwt: false
with_linear: false
with_mel: false
with_spk_embed: false
with_w2v2_feat: false
with_wav: true
binarizer_cls: data_gen.rosvot_binarizer.RosvotBinarizer
binary_data_dir: data/binary/m4
bkb_layers: 2
bkb_net: conformer
channel_multiples: 1-1-1
check_val_every_n_epoch: 10
clip_grad_norm: 1
clip_grad_value: 0
conformer_kernel: 9
dataset_downsample_rate: 1.0
debug: false
dropout: 0.1
ds_names: m4
ds_names_in_testing: ''
ds_names_in_training: ''
ds_workers: 8
endless_ds: true
eval_max_batches: -1
f0_add_noise: gaussian:0.04
f0_bin: 512
f0_filepath: ''
f0_max: 900
f0_min: 50
fft_size: 512
find_unused_parameters: true
fmax: 12000
fmin: 30
frames_multiple: 8
gen_dir_name: ''
hidden_size: 256
hop_size: 128
infer_meta_path: ''
infer_print_skipped: true
infer_regulate_real_note_itv: true
input_process_name: none
label_pos_weight_decay: 0.95
lambda_note_bd: 1.0
lambda_note_bd_focal: 3.0
lambda_note_bd_slur_punish: 0.0
lambda_note_pitch: 1.0
lambda_word_bd: 1.0
lambda_word_bd_focal: 3.0
load_ckpt: ''
loud_norm: false
lr: 5.0e-06
max_epochs: 1000
max_frames: 4000
max_input_tokens: 1550
max_sentences: 128
max_tokens: 80000
max_updates: 40000
max_valid_sentences: 1
max_valid_tokens: 60000
mel_add_noise: gaussian:0.05
mel_vmax: 1.5
mel_vmin: -6
min_frames: 0
min_note_dur: 80
min_word_dur: 20
model: rosvot
noise_in_test: false
noise_prob: 0.8
noise_snr: 6-20
note_bd_add_noise: gaussian:0.002
note_bd_focal_loss: 5.0
note_bd_min_gap: 90
note_bd_ratio: 2.42312
note_bd_ref_min_gap: 40
note_bd_start: 0
note_bd_temperature: 0.2
note_bd_threshold: 0.8
note_num: 85
note_pitch_label_smoothing: 0.005
note_pitch_start: 0
note_pitch_temperature: 0.01
note_start: 30
num_ckpt_keep: 3
num_sanity_val_steps: 5
num_valid_plots: 10
num_valid_stats: 100
optimizer_adam_beta1: 0.9
optimizer_adam_beta2: 0.98
out_wav_norm: false
pe: rmvpe
pe_ckpt: checkpoints/rmvpe/model.pt
pin_memory: true
pitch_attn_num_head: 4
pitch_type: frame
print_nan_grads: false
processed_data_dir: data/processed/m4
profile_infer: false
raw_data_dir: ''
rename_tmux: false
resume_from_checkpoint: 0
save_best: true
save_codes:
- modules
- tasks
save_f0: false
save_gt: true
save_plot: true
scheduler: step_lr
scheduler_lr_gamma: 0.998
scheduler_lr_step_size: 500
seed: 42
soft_note_bd_func: gaussian:80
soft_word_bd_func: gaussian:80
sort_by_len: true
task_cls: tasks.rosvot.task.RobustWordbdTask
tb_log_interval: 100
test_ids: []
test_input_yaml: ''
test_set_name: test
train_set_name: train
train_sets: ''
unet_skip_layer: false
updown_rates: 2-2-2
use_mel: true
use_mel_bins: 40
use_pitch_embed: true
use_soft_note: false
use_soft_note_bd: true
use_soft_word_bd: true
use_spk_embed: false
use_spk_id: false
use_wav: false
use_word_input: false
val_check_interval: 500
valid_infer_interval: 10000
valid_monitor_key: val_loss
valid_monitor_mode: min
valid_set_name: valid
warmup_updates: 0
weight_decay: 0
win_size: 512
word_bd_add_noise: gaussian:0.002
word_bd_focal_loss: 5.0
word_bd_min_gap: 90
word_bd_ratio: 2.2
word_bd_start: 0
word_bd_temperature: 0.2
word_bd_threshold: 0.9
work_dir: checkpoints/240613-rwbd-03