|
|
project_root: "." |
|
|
evaluation: |
|
|
checkpoint_path: "" |
|
|
output_dir: "outputs" |
|
|
test_set_path: "inputs/input.json" |
|
|
negative_style_prompt: ${project_root}/public/vocal.npy |
|
|
num_samples: null |
|
|
batch_size: 1 |
|
|
random_crop_style: false |
|
|
vae_type: 'diffrhythm' |
|
|
num_style_secs: 30 |
|
|
ignore_style: false |
|
|
use_prompt_style: false |
|
|
|
|
|
dataset: |
|
|
pattern: "placeholder" |
|
|
shuffle: false |
|
|
resample_by_duration_threshold: null |
|
|
always_crop_from_beginning: true |
|
|
always_use_style_index: 0 |
|
|
|
|
|
sample_kwargs: |
|
|
cfg_range: |
|
|
- 0.05 |
|
|
- 1 |
|
|
dual_cfg: |
|
|
- 4.7 |
|
|
- 2.5 |
|
|
steps: 50 |
|
|
|
|
|
model: |
|
|
num_channels: 64 |
|
|
cfm: |
|
|
max_frames: ${max_frames} |
|
|
num_channels: ${model.num_channels} |
|
|
dual_drop_prob: [0.1, 0.5] |
|
|
no_edit: true |
|
|
|
|
|
dit: |
|
|
max_frames: ${max_frames} |
|
|
mel_dim: ${model.num_channels} |
|
|
dim: 1408 |
|
|
depth: 16 |
|
|
heads: 32 |
|
|
ff_mult: 4 |
|
|
text_dim: 512 |
|
|
conv_layers: 4 |
|
|
grad_ckpt: true |
|
|
use_implicit_duration: true |
|
|
|
|
|
data: |
|
|
train_dataset: |
|
|
max_frames: ${max_frames} |
|
|
multiple_styles: true |
|
|
sampling_rate: 44100 |
|
|
shuffle: true |
|
|
silence_latent_path: ${project_root}/public/silience_latent.pt |
|
|
tokenizer_path: ${project_root}/public/en_us_cmudict_ipa_forward.pt |
|
|
lrc_upsample_factor: ${lrc_upsample_factor} |
|
|
filler: average_sparse |
|
|
phonemizer_checkpoint: ${project_root}/public/en_us_cmudict_ipa_forward.pt |
|
|
|
|
|
|
|
|
max_frames: 5000 |
|
|
lrc_upsample_factor: 4 |
|
|
seed: 42 |
