datnt-castalk commited on
Commit
8410815
·
verified ·
1 Parent(s): 84413ff

Upload folder using huggingface_hub

Browse files
Files changed (3) hide show
  1. epoch_2nd_00039.pth +3 -0
  2. gackt.yml +20 -0
  3. train.log +0 -0
epoch_2nd_00039.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e220ad551da153fe17a53b3472f5d83dbe48559c238735e05a750daa20918082
3
+ size 3647533280
gackt.yml ADDED
@@ -0,0 +1,20 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {ASR_config: Utils/ASR/config.yml, ASR_path: Utils/ASR/bst_00080.pth, F0_path: Utils/JDC/bst.t7,
2
+ PLBERT_dir: Utils/PLBERT/, batch_size: 3, data_params: {OOD_data: Data/OOD_LargeScale_.csv,
3
+ min_length: 50, root_path: /vast/tts/share/gackt_processed/segments/, train_data: Data/GACKT_CD.txt,
4
+ val_data: Data/val.txt}, device: cuda, epochs: 50, load_only_params: true, log_dir: Models/GACKT_Finetune_24KHZ_LLR,
5
+ log_interval: 10, loss_params: {TMA_epoch: 9, diff_epoch: 0, joint_epoch: 30, lambda_F0: 1.0,
6
+ lambda_ce: 20.0, lambda_diff: 1.0, lambda_dur: 1.0, lambda_gen: 1.0, lambda_mel: 10.0,
7
+ lambda_mono: 1.0, lambda_norm: 1.0, lambda_s2s: 1.0, lambda_slm: 1.0, lambda_sty: 1.0},
8
+ max_len: 1500, model_params: {decoder: {gen_istft_hop_size: 5, gen_istft_n_fft: 20,
9
+ resblock_dilation_sizes: [[1, 3, 5], [1, 3, 5], [1, 3, 5]], resblock_kernel_sizes: [
10
+ 3, 7, 11], type: istftnet, upsample_initial_channel: 512, upsample_kernel_sizes: [
11
+ 20, 12], upsample_rates: [10, 6]}, diffusion: {dist: {estimate_sigma_data: true,
12
+ mean: -3.0, sigma_data: 0.19918164490248824, std: 1.0}, embedding_mask_proba: 0.1,
13
+ transformer: {head_features: 64, multiplier: 2, num_heads: 8, num_layers: 3}},
14
+ dim_in: 64, dropout: 0.2, hidden_dim: 512, max_conv_dim: 512, max_dur: 50, multispeaker: true,
15
+ n_layer: 3, n_mels: 80, n_token: 178, slm: {hidden: 1280, initial_channel: 64,
16
+ model: Respair/Whisper_Large_v2_Encoder_Block, nlayers: 33, sr: 16000}, style_dim: 128},
17
+ optimizer_params: {bert_lr: 4.0e-06, ft_lr: 4.0e-06, lr: 4.0e-05}, preprocess_params: {
18
+ spect_params: {hop_length: 300, n_fft: 2048, win_length: 1200}, sr: 24000}, pretrained_model: Models/Style_Tsukasa_v02/Top_ckpt_24khz.pth,
19
+ save_freq: 5, second_stage_load_pretrained: true, slmadv_params: {batch_percentage: 0.5,
20
+ iter: 20, max_len: 500, min_len: 400, scale: 0.01, sig: 1.5, thresh: 5}}
train.log ADDED
The diff for this file is too large to render. See raw diff