more code refactor

GPT_SoVITS/configs/s1.yaml

@@ -1,31 +1,31 @@
 train:
-    seed: 1234
-    epochs: 300
-    batch_size: 8
-    gradient_accumulation: 4
-    save_every_n_epoch: 1
-    precision: 16
-    gradient_clip: 1.0
+  seed: 1234
+  epochs: 300
+  batch_size: 8
+  gradient_accumulation: 4
+  save_every_n_epoch: 1
+  precision: 16
+  gradient_clip: 1.0
 optimizer:
-    lr: 0.01
-    lr_init: 0.00001
-    lr_end: 0.0001
-    warmup_steps: 2000
-    decay_steps: 40000
+  lr: 0.01
+  lr_init: 0.00001
+  lr_end: 0.0001
+  warmup_steps: 2000
+  decay_steps: 40000
 data:
-    max_eval_sample: 8
-    max_sec: 54
-    num_workers: 1
-    pad_val: 1024 # same with EOS in model
+  max_eval_sample: 8
+  max_sec: 54
+  num_workers: 1
+  pad_val: 1024 # same with EOS in model
 model:
-    vocab_size: 1025
-    phoneme_vocab_size: 512
-    embedding_dim: 512
-    hidden_dim: 512
-    head: 16
-    linear_units: 2048
-    n_layer: 12
-    dropout: 0
-    EOS: 1024
+  vocab_size: 1025
+  phoneme_vocab_size: 512
+  embedding_dim: 512
+  hidden_dim: 512
+  head: 16
+  linear_units: 2048
+  n_layer: 12
+  dropout: 0
+  EOS: 1024
 inference:
-    top_k: 5
+  top_k: 5

GPT_SoVITS/configs/s1big.yaml

@@ -1,31 +1,31 @@
 train:
-    seed: 1234
-    epochs: 300
-    batch_size: 8
-    gradient_accumulation: 4
-    save_every_n_epoch: 1
-    precision: 16-mixed
-    gradient_clip: 1.0
+  seed: 1234
+  epochs: 300
+  batch_size: 8
+  gradient_accumulation: 4
+  save_every_n_epoch: 1
+  precision: 16-mixed
+  gradient_clip: 1.0
 optimizer:
-    lr: 0.01
-    lr_init: 0.00001
-    lr_end: 0.0001
-    warmup_steps: 2000
-    decay_steps: 40000
+  lr: 0.01
+  lr_init: 0.00001
+  lr_end: 0.0001
+  warmup_steps: 2000
+  decay_steps: 40000
 data:
-    max_eval_sample: 8
-    max_sec: 54
-    num_workers: 1
-    pad_val: 1024 # same with EOS in model
+  max_eval_sample: 8
+  max_sec: 54
+  num_workers: 1
+  pad_val: 1024 # same with EOS in model
 model:
-    vocab_size: 1025
-    phoneme_vocab_size: 512
-    embedding_dim: 1024
-    hidden_dim: 1024
-    head: 16
-    linear_units: 2048
-    n_layer: 16
-    dropout: 0
-    EOS: 1024
+  vocab_size: 1025
+  phoneme_vocab_size: 512
+  embedding_dim: 1024
+  hidden_dim: 1024
+  head: 16
+  linear_units: 2048
+  n_layer: 16
+  dropout: 0
+  EOS: 1024
 inference:
-    top_k: 5
+  top_k: 5

GPT_SoVITS/configs/s1big2.yaml

@@ -1,31 +1,31 @@
 train:
-    seed: 1234
-    epochs: 300
-    batch_size: 12
-    gradient_accumulation: 4
-    save_every_n_epoch: 1
-    precision: 16-mixed
-    gradient_clip: 1.0
+  seed: 1234
+  epochs: 300
+  batch_size: 12
+  gradient_accumulation: 4
+  save_every_n_epoch: 1
+  precision: 16-mixed
+  gradient_clip: 1.0
 optimizer:
-    lr: 0.01
-    lr_init: 0.00001
-    lr_end: 0.0001
-    warmup_steps: 2000
-    decay_steps: 40000
+  lr: 0.01
+  lr_init: 0.00001
+  lr_end: 0.0001
+  warmup_steps: 2000
+  decay_steps: 40000
 data:
-    max_eval_sample: 8
-    max_sec: 54
-    num_workers: 1
-    pad_val: 1024 # same with EOS in model
+  max_eval_sample: 8
+  max_sec: 54
+  num_workers: 1
+  pad_val: 1024 # same with EOS in model
 model:
-    vocab_size: 1025
-    phoneme_vocab_size: 512
-    embedding_dim: 1024
-    hidden_dim: 1024
-    head: 16
-    linear_units: 2048
-    n_layer: 6
-    dropout: 0
-    EOS: 1024
+  vocab_size: 1025
+  phoneme_vocab_size: 512
+  embedding_dim: 1024
+  hidden_dim: 1024
+  head: 16
+  linear_units: 2048
+  n_layer: 6
+  dropout: 0
+  EOS: 1024
 inference:
-    top_k: 5
+  top_k: 5

GPT_SoVITS/configs/s1longer.yaml

@@ -1,31 +1,31 @@
 train:
-    seed: 1234
-    epochs: 20
-    batch_size: 8
-    save_every_n_epoch: 1
-    precision: 16-mixed
-    gradient_clip: 1.0
+  seed: 1234
+  epochs: 20
+  batch_size: 8
+  save_every_n_epoch: 1
+  precision: 16-mixed
+  gradient_clip: 1.0
 optimizer:
-    lr: 0.01
-    lr_init: 0.00001
-    lr_end: 0.0001
-    warmup_steps: 2000
-    decay_steps: 40000
+  lr: 0.01
+  lr_init: 0.00001
+  lr_end: 0.0001
+  warmup_steps: 2000
+  decay_steps: 40000
 data:
-    max_eval_sample: 8
-    max_sec: 54
-    num_workers: 4
-    pad_val: 1024 # same with EOS in model
+  max_eval_sample: 8
+  max_sec: 54
+  num_workers: 4
+  pad_val: 1024 # same with EOS in model
 model:
-    vocab_size: 1025
-    phoneme_vocab_size: 512
-    embedding_dim: 512
-    hidden_dim: 512
-    head: 16
-    linear_units: 2048
-    n_layer: 24
-    dropout: 0
-    EOS: 1024
-    random_bert: 0
+  vocab_size: 1025
+  phoneme_vocab_size: 512
+  embedding_dim: 512
+  hidden_dim: 512
+  head: 16
+  linear_units: 2048
+  n_layer: 24
+  dropout: 0
+  EOS: 1024
+  random_bert: 0
 inference:
-    top_k: 5
+  top_k: 5

GPT_SoVITS/configs/s1mq.yaml

@@ -1,77 +1,77 @@
 train:
-    seed: 1234
-    epochs: 100
-    batch_size: 6
-    gradient_accumulation: 4
-    save_every_n_epoch: 1
-    precision: 32
-    gradient_clip: 1.0
+  seed: 1234
+  epochs: 100
+  batch_size: 6
+  gradient_accumulation: 4
+  save_every_n_epoch: 1
+  precision: 32
+  gradient_clip: 1.0
 optimizer:
-    lr: 0.01
-    lr_init: 0.00001
-    lr_end: 0.0001
-    warmup_steps: 2000
-    decay_steps: 40000
+  lr: 0.01
+  lr_init: 0.00001
+  lr_end: 0.0001
+  warmup_steps: 2000
+  decay_steps: 40000
 data:
-    max_eval_sample: 8
-    max_sec: 40
-    num_workers: 1
-    pad_val: 1024 # same with EOS in model
+  max_eval_sample: 8
+  max_sec: 40
+  num_workers: 1
+  pad_val: 1024 # same with EOS in model
 model:
-    saving_path: "ckpt/"
-    resume_checkpoint: null
-    vocoder_config_path: "quantizer/new_ckpt/config.json"
-    vocoder_ckpt_path: "quantizer/new_ckpt/g_00600000"
-    datadir: "/home/liweiche/GigaSpeech/wavs"
-    metapath: "/home/liweiche/GigaSpeech/train2.json"
-    val_metapath: "/home/liweiche/GigaSpeech/dev2.json"
-    sampledir: "logs/"
-    pretrained_path: null
-    lr: 0.0001
-    batch_size: 200.0
-    train_bucket_size: 8192
-    training_step: 800000
-    optim_flat_percent: 0.0
-    warmup_step: 50
-    adam_beta1: 0.9
-    adam_beta2: 0.98
-    ffd_size: 3072
-    hidden_size: 768
-    enc_nlayers: 6
-    dec_nlayers: 6
-    nheads: 12
-    ar_layer: 4
-    ar_ffd_size: 1024
-    ar_hidden_size: 256
-    ar_nheads: 4
-    aligner_softmax_temp: 1.0
-    layer_norm_eps: 0.00001
-    speaker_embed_dropout: 0.05
-    label_smoothing: 0.0
-    val_check_interval: 5000
-    check_val_every_n_epoch: 1
-    precision: "fp16"
-    nworkers: 16
-    distributed: true
-    accelerator: "ddp"
-    version: null
-    accumulate_grad_batches: 1
-    use_repetition_token: true
-    use_repetition_gating: false
-    repetition_penalty: 1.0
-    sampling_temperature: 1.0
-    top_k: -1
-    min_top_k: 3
-    top_p: 0.8
-    sample_num: 4
-    length_penalty_max_length: 15000
-    length_penalty_max_prob: 0.95
-    max_input_length: 2048
-    max_output_length: 2000
-    sample_rate: 16000
-    n_codes: 1024
-    n_cluster_groups: 1
-    phone_context_window: 4
-    phoneset_size: 1000
+  saving_path: "ckpt/"
+  resume_checkpoint: null
+  vocoder_config_path: "quantizer/new_ckpt/config.json"
+  vocoder_ckpt_path: "quantizer/new_ckpt/g_00600000"
+  datadir: "/home/liweiche/GigaSpeech/wavs"
+  metapath: "/home/liweiche/GigaSpeech/train2.json"
+  val_metapath: "/home/liweiche/GigaSpeech/dev2.json"
+  sampledir: "logs/"
+  pretrained_path: null
+  lr: 0.0001
+  batch_size: 200.0
+  train_bucket_size: 8192
+  training_step: 800000
+  optim_flat_percent: 0.0
+  warmup_step: 50
+  adam_beta1: 0.9
+  adam_beta2: 0.98
+  ffd_size: 3072
+  hidden_size: 768
+  enc_nlayers: 6
+  dec_nlayers: 6
+  nheads: 12
+  ar_layer: 4
+  ar_ffd_size: 1024
+  ar_hidden_size: 256
+  ar_nheads: 4
+  aligner_softmax_temp: 1.0
+  layer_norm_eps: 0.00001
+  speaker_embed_dropout: 0.05
+  label_smoothing: 0.0
+  val_check_interval: 5000
+  check_val_every_n_epoch: 1
+  precision: "fp16"
+  nworkers: 16
+  distributed: true
+  accelerator: "ddp"
+  version: null
+  accumulate_grad_batches: 1
+  use_repetition_token: true
+  use_repetition_gating: false
+  repetition_penalty: 1.0
+  sampling_temperature: 1.0
+  top_k: -1
+  min_top_k: 3
+  top_p: 0.8
+  sample_num: 4
+  length_penalty_max_length: 15000
+  length_penalty_max_prob: 0.95
+  max_input_length: 2048
+  max_output_length: 2000
+  sample_rate: 16000
+  n_codes: 1024
+  n_cluster_groups: 1
+  phone_context_window: 4
+  phoneset_size: 1000
 inference:
-    top_k: 5
+  top_k: 5

GPT_SoVITS/configs/train.yaml

@@ -1,32 +1,32 @@
 gpu:
-    n_card: 1
-    n_process_per_card: 2
+  n_card: 1
+  n_process_per_card: 2
 io:
-    text_path: D:\RVC1006\GPT-SoVITS\GPT_SoVITS
-    save_every_n_epoch: 1
-    precision: 16-mixed
-    gradient_clip: 1.0
+  text_path: D:\RVC1006\GPT-SoVITS\GPT_SoVITS
+  save_every_n_epoch: 1
+  precision: 16-mixed
+  gradient_clip: 1.0
 optimizer:
-    lr: 0.01
-    lr_init: 0.00001
-    lr_end: 0.0001
-    warmup_steps: 2000
-    decay_steps: 40000
+  lr: 0.01
+  lr_init: 0.00001
+  lr_end: 0.0001
+  warmup_steps: 2000
+  decay_steps: 40000
 data:
-    max_eval_sample: 8
-    max_sec: 54
-    num_workers: 1
-    pad_val: 1024 # same with EOS in model
+  max_eval_sample: 8
+  max_sec: 54
+  num_workers: 1
+  pad_val: 1024 # same with EOS in model
 model:
-    vocab_size: 1025
-    phoneme_vocab_size: 512
-    embedding_dim: 512
-    hidden_dim: 512
-    head: 16
-    linear_units: 2048
-    n_layer: 24
-    dropout: 0
-    EOS: 1024
-    random_bert: 0
+  vocab_size: 1025
+  phoneme_vocab_size: 512
+  embedding_dim: 512
+  hidden_dim: 512
+  head: 16
+  linear_units: 2048
+  n_layer: 24
+  dropout: 0
+  EOS: 1024
+  random_bert: 0
 inference:
-    top_k: 5
+  top_k: 5