accumulate_grad_batches: 1
add_word_pos: true
amp: true
ast_context_model: MIT/ast-finetuned-audioset-10-10-0.4593
ast_context_scale: 0.25
ast_context_valid_prefix: false
audio_num_mel_bins: 100
audio_sample_rate: 24000
base_config:
- ./agent-1_nonlinear_phone_teacher_5k.yaml
binarizer_cls: data_gen.tts.base_binarizer.BaseBinarizer
binary_data_dir: /mnt/DATA/ResearchDrive/audiopaint/data/binary/libritts_bigvgan
cgpc_gt_view: utt
cgpc_tau: 0.1
clip_grad_norm: 1
clip_grad_value: 0
cond_dim: 0
dec_dilations:
- 1
- 1
- 1
- 1
dec_ffn_kernel_size: 9
dec_inp_add_noise: false
dec_kernel_size: 5
dec_layers: 4
dec_post_net_kernel: 3
decoder_rnn_dim: 0
decoder_type: conv
deletion_source_duration_enabled: false
deletion_source_probability: 0.0
detach_postflow_input: true
diff_decoder_type: wavenet
diff_loss_type: l1
dilation_cycle_length: 1
donor_splice_attempts: 24
donor_splice_exclude_conn_loss: true
donor_splice_match_prosody: true
donor_splice_neighbor_search: 128
donor_splice_operation: insertion
donor_splice_probability: 0.5
donor_splice_seam_top_k: 8
donor_splice_target_phone_max: 9
donor_splice_target_phone_min: 3
dropout: 0.0
ds_name: libritts
ds_workers: 2
dur_level: word
dur_predictor_kernel: 5
dur_predictor_layers: 3
enc_dec_norm: ln
enc_dilations:
- 1
- 1
- 1
- 1
enc_ffn_kernel_size: 5
enc_kernel_size: 5
enc_layers: 4
enc_post_net_kernel: 3
enc_pre_ln: true
enc_prenet: true
encoder_K: 8
encoder_type: conv
endless_ds: true
eval_max_batches: -1
f0_max: 600
f0_min: 80
ffn_hidden_size: 768
fft_size: 1024
fmax: null
fmin: 0
frames_multiple: 1
gen_dir_name: ''
gst_seed: 1234
gst_trainable: false
hidden_size: 192
hlac_frame_metric: norm
hlac_levels: fpw
hlac_w_frame: 1.0
hlac_w_phone: 1.0
hlac_w_word: 1.0
hop_size: 256
infer_mask_ratio: 0.3
keep_bins: 100
kl_min: 0.0
kl_start_steps: 10000
lambda_cgpc: 0.0
lambda_deletion_source_duration: 0.0
lambda_energy: 0.1
lambda_f0: 1.0
lambda_hlac: 0.0
lambda_kl: 1.0
lambda_mel_adv: 0.05
lambda_nonlinear_phone: 0.01
lambda_ph_dur: 0.1
lambda_phone_mel_contrast: 0.05
lambda_sent_dur: 0.0
lambda_uv: 1.0
lambda_word_dur: 1.0
latent_size: 16
layers_in_block: 2
load_ckpt: ''
load_ckpt_min_match: 0.9
loud_norm: false
lr: 2.0e-05
mask_type: continuous_
max_beta: 0.06
max_frames: 1548
max_input_tokens: 1550
max_sentences: 16
max_tokens: 40000
max_updates: 5000
max_valid_sentences: 1
max_valid_tokens: 60000
mel_disc_hidden_size: 128
mel_losses: l1:0.5|ssim:0.5
mel_type: bigvgan
mel_vmax: 1.5
mel_vmin: -6
min_frames: 0
noise_scale: 0.8
nonlinear_phone_teacher: assets/agent-1_nonlinear_phone_teacher.pt
num_ckpt_keep: 12
num_heads: 2
num_sanity_val_steps: 5
num_spk: 1261
num_valid_plots: 10
optimizer_adam_beta1: 0.9
optimizer_adam_beta2: 0.98
out_wav_norm: false
phone_mel_prototypes: assets/agent-1_phone_mel_prototypes.npz
phone_mel_temperature: 0.1
pitch_extractor: parselmouth
pitch_type: frame
posterior_start_steps: 0
predictor_dropout: 0.2
predictor_grad: 0.1
predictor_hidden: -1
predictor_kernel: 5
predictor_layers: 5
print_nan_grads: false
prior_flow_hidden: 64
prior_flow_kernel_size: 3
prior_flow_n_blocks: 4
processed_data_dir: /mnt/DATA/ResearchDrive/audiopaint/data/processed/libritts
profile_infer: false
raw_data_dir: data/raw/libritts
residual_channels: 256
residual_layers: 20
resume_from_checkpoint: 0
rope_half: true
rope_reset_at_nt: true
save_best: false
save_codes:
- tasks
- modules
save_gt: true
schedule_type: vpsde
scheduler: warmup
seed: 1234
share_wn_layers: 4
sigmoid_scale: false
sort_by_len: true
spec_max:
- 1.0604
- 1.4571
- 1.2852
- 1.3722
- 1.1957
- 1.6533
- 1.5299
- 1.1385
- 1.2986
- 1.3136
- 1.5607
- 1.3109
- 1.7511
- 1.7297
- 1.5513
- 1.6957
- 1.3587
- 1.8329
- 1.4253
- 1.2918
- 1.4762
- 1.3411
- 1.4748
- 1.3285
- 1.4044
- 1.1602
- 1.4003
- 1.2423
- 1.2876
- 1.0298
- 1.2514
- 1.1693
- 1.2147
- 1.1453
- 1.055
- 1.2303
- 1.4054
- 1.0691
- 1.1672
- 1.21
- 0.6927
- 0.6037
- 0.9786
- 0.361
- 0.7867
- 0.3277
- 0.4398
- 0.5989
- 0.4387
- 0.3292
- 0.1331
- 0.2449
- 0.2704
- -0.1298
- 0.0984
- -0.1073
- -0.1773
- -0.1167
- -0.1131
- -0.1729
- -0.3771
- -0.3133
- -0.5095
- -0.6342
- -0.1225
- -0.0924
- -0.2996
- -0.1631
- -0.4548
- -0.6299
- -0.4956
- -0.2327
- -0.2719
- -0.8649
- -0.4402
- -0.6376
- -0.7074
- -0.552
- -0.7659
- -0.7273
- -0.8326
- -0.7609
- -0.9281
- -0.6764
- -0.3611
- -1.0908
- -1.167
- -0.9377
- -0.699
- -0.7402
- -1.1614
- -1.1782
- -0.6824
- -0.8936
- -0.7902
- -1.1805
- -1.3148
- -1.3482
- -1.61
- -2.2738
spec_min:
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
- -11.5129
task_cls: tasks.speech_editing.spec_denoiser.SpeechDenoiserTask
tb_log_interval: 100
test_ids: []
test_set_name: test
text_encoder_postnet: true
timescale: 1
timesteps: 8
train_set_name: train
train_sets: ''
training_mask_ratio: 0.15
trunk_adaptation: audio_lora
trunk_donor: Qwen/Qwen2.5-0.5B
trunk_freeze_tok_emb: true
trunk_grad_ckpt: true
trunk_inject: add
trunk_layer_pick: ends
trunk_layers: 14
trunk_lora_alpha: 16.0
trunk_lora_rank: 8
trunk_max_text: 48
trunk_pretrained_lr_scale: 0.05
trunk_use_ph: true
trunk_use_text: true
trunk_warm_start: true
two_stage: true
use_ast_context: false
use_cond_proj: false
use_conn_loss: true
use_dualffn_trunk: true
use_energy_embed: false
use_gt_dur: false
use_gt_f0: false
use_latent_cond: false
use_perceptual_loss: true
use_pitch_embed: true
use_post_flow: true
use_prior_flow: true
use_spk_embed: true
use_spk_id: false
use_txt_cond: true
use_uv: true
val_check_interval: 500
valid_infer_interval: 2000
valid_monitor_key: val_loss
valid_monitor_mode: min
valid_set_name: valid
vocoder: BigVGAN
vocoder_ckpt: pretrained/hifigan_hifitts
warmup_updates: 1
weight_decay: 0
win_size: 1024
work_dir: /mnt/DATA/ResearchDrive/audiopaint//checkpoints/agent-2_matched_horizon_no_deletion_5k