Add files using upload-large-folder tool

WavCube-pro/checkpoints/vocos_checkpoint_epoch=34_step=200000_val_loss=3.2140.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3be60dc837b1c90f4717d4b87a0616295785087b683212b27d8883ef1e94f684
+size 11575027210

WavCube-pro/config.yaml

@@ -0,0 +1,160 @@
+# pytorch_lightning==1.8.6
+seed_everything: 4444
+trainer:
+  logger:
+    class_path: pytorch_lightning.loggers.TensorBoardLogger
+    init_args:
+      save_dir: logs/wavlmvae-mimo-librispeech-stage2_kl1e-4_ae_300mdeco_6k
+      name: first
+      version: null
+      log_graph: false
+      default_hp_metric: true
+      prefix: ''
+      sub_dir: null
+      logdir: null
+      comment: ''
+      purge_step: null
+      max_queue: 10
+      flush_secs: 120
+      filename_suffix: ''
+      write_to_disk: true
+      comet_config:
+        disabled: true
+  enable_checkpointing: true
+  callbacks:
+  - class_path: pytorch_lightning.callbacks.LearningRateMonitor
+    init_args:
+      logging_interval: null
+      log_momentum: false
+  - class_path: pytorch_lightning.callbacks.ModelSummary
+    init_args:
+      max_depth: 2
+  - class_path: pytorch_lightning.callbacks.ModelCheckpoint
+    init_args:
+      dirpath: null
+      filename: vocos_checkpoint_{epoch}_{step}_{val_loss:.4f}
+      monitor: val_loss
+      verbose: false
+      save_last: true
+      save_top_k: -1
+      save_weights_only: false
+      mode: min
+      auto_insert_metric_name: true
+      every_n_train_steps: 1000
+      train_time_interval: null
+      every_n_epochs: null
+      save_on_train_epoch_end: null
+  - class_path: vocos.helpers.GradNormCallback
+  default_root_dir: null
+  gradient_clip_val: null
+  gradient_clip_algorithm: null
+  num_nodes: 8
+  num_processes: null
+  devices: '8'
+  gpus: null
+  auto_select_gpus: false
+  tpu_cores: null
+  ipus: null
+  enable_progress_bar: true
+  overfit_batches: 0.0
+  track_grad_norm: -1
+  check_val_every_n_epoch: 1
+  fast_dev_run: false
+  accumulate_grad_batches: null
+  max_epochs: null
+  min_epochs: null
+  max_steps: 1000000
+  min_steps: null
+  max_time: null
+  limit_train_batches: null
+  limit_val_batches: 100
+  limit_test_batches: null
+  limit_predict_batches: null
+  val_check_interval: null
+  log_every_n_steps: 50
+  accelerator: gpu
+  strategy: ddp
+  sync_batchnorm: false
+  precision: 32
+  enable_model_summary: true
+  num_sanity_val_steps: 2
+  resume_from_checkpoint: logs/wavlmvae-mimo-librispeech-stage2_kl1e-4_ae_300mdeco_6k/first/version_2/checkpoints/vocos_checkpoint_epoch=11_step=70000_val_loss=3.5162.ckpt
+  profiler: null
+  benchmark: null
+  deterministic: null
+  reload_dataloaders_every_n_epochs: 0
+  auto_lr_find: false
+  replace_sampler_ddp: true
+  detect_anomaly: false
+  auto_scale_batch_size: false
+  plugins: null
+  amp_backend: native
+  amp_level: null
+  move_metrics_to_cpu: false
+  multiple_trainloader_mode: max_size_cycle
+  inference_mode: true
+model:
+  class_path: vocos.experiment.MiMoWavLMVAEExp
+  init_args:
+    feature_extractor:
+      class_path: vocos.feature_extractors.WavLMVAEFeatures
+      init_args:
+        model_id: ckpts/wavlm-large
+        layer_idx: -1
+        freeze_model: true
+        latent_dim: 128
+        stage: 2
+        stage1_ckpt_path: logs/wavlmvae-mimo-librispeech-stage1_kl1e-4_ae_300mdeco_6k/first/version_2/checkpoints/vocos_checkpoint_epoch=41_step=138000_val_loss=6.2627.ckpt
+        use_vae: false
+        use_sigma_vae: false
+        use_temporal_downsampling: false
+        apply_mask: false
+        mask_time_prob: 0.15
+        mask_time_length: 10
+    backbone:
+      class_path: vocos.models.MiMoBackbone
+      init_args:
+        d_model: 1024
+        decoder_attention_heads: 16
+        decoder_ffn_dim: 4096
+        sampling_rate: 16000
+        hop_length: 160
+        window_size: 640
+        nfft: 640
+        upsample: true
+        latent_dim: 128
+        decoder_layers: 24
+    head: null
+    sample_rate: 16000
+    initial_learning_rate: 0.0001
+    num_warmup_steps: 5000
+    mel_loss_coeff: 4.5
+    mrd_loss_coeff: 1.0
+    kl_loss_coeff: 0.0001
+    sr_loss_coeff: 1.0
+    gan_loss_coeff: 0.1
+    pretrain_mel_steps: 0
+    decay_mel_coeff: false
+    evaluate_utmos: true
+    evaluate_pesq: true
+    evaluate_periodicty: false
+    evaluate_stoi: false
+    evaluate_pesq_wb: false
+    evaluate_sim: true
+data:
+  class_path: vocos.dataset.VocosEmiliaDataModule
+  init_args:
+    train_params:
+      filelist_path: data/librispeech_train_librilight_small_meidum_6k
+      sampling_rate: 16000
+      num_samples: 160000
+      batch_size: 8
+      num_workers: 8
+    val_params:
+      filelist_path: data/librispeech_test_clean
+      sampling_rate: 16000
+      num_samples: 160000
+      batch_size: 8
+      num_workers: 8
+optimizer: null
+lr_scheduler: null

WavCube-pro/hparams.yaml

@@ -0,0 +1,16 @@
+sample_rate: 16000
+initial_learning_rate: 0.0001
+num_warmup_steps: 5000
+mel_loss_coeff: 4.5
+mrd_loss_coeff: 1.0
+pretrain_mel_steps: 0
+decay_mel_coeff: false
+evaluate_utmos: true
+evaluate_pesq: true
+evaluate_periodicty: false
+evaluate_stoi: false
+evaluate_pesq_wb: false
+evaluate_sim: true
+kl_loss_coeff: 0.0001
+sr_loss_coeff: 1.0
+gan_loss_coeff: 0.1

WavCube/checkpoints/vocos_checkpoint_epoch=177_step=195000_val_loss=3.3080.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ba70c7d19271036b1a2cc5efaba3470fe81eced7b9cff3803515032831f5d609
+size 11574970721

WavCube/config.yaml

@@ -0,0 +1,157 @@
+# pytorch_lightning==1.8.6
+seed_everything: 4444
+trainer:
+  logger:
+    class_path: pytorch_lightning.loggers.TensorBoardLogger
+    init_args:
+      save_dir: logs/wavlmvae-mimo-librispeech-stage2_kl1e-4_ae_300mdeco
+      name: first
+      version: null
+      log_graph: false
+      default_hp_metric: true
+      prefix: ''
+      sub_dir: null
+      logdir: null
+      comment: ''
+      purge_step: null
+      max_queue: 10
+      flush_secs: 120
+      filename_suffix: ''
+      write_to_disk: true
+      comet_config:
+        disabled: true
+  enable_checkpointing: true
+  callbacks:
+  - class_path: pytorch_lightning.callbacks.LearningRateMonitor
+    init_args:
+      logging_interval: null
+      log_momentum: false
+  - class_path: pytorch_lightning.callbacks.ModelSummary
+    init_args:
+      max_depth: 2
+  - class_path: pytorch_lightning.callbacks.ModelCheckpoint
+    init_args:
+      dirpath: null
+      filename: vocos_checkpoint_{epoch}_{step}_{val_loss:.4f}
+      monitor: val_loss
+      verbose: false
+      save_last: true
+      save_top_k: -1
+      save_weights_only: false
+      mode: min
+      auto_insert_metric_name: true
+      every_n_train_steps: 5000
+      train_time_interval: null
+      every_n_epochs: null
+      save_on_train_epoch_end: null
+  - class_path: vocos.helpers.GradNormCallback
+  default_root_dir: null
+  gradient_clip_val: null
+  gradient_clip_algorithm: null
+  num_nodes: 8
+  num_processes: null
+  devices: '8'
+  gpus: null
+  auto_select_gpus: false
+  tpu_cores: null
+  ipus: null
+  enable_progress_bar: true
+  overfit_batches: 0.0
+  track_grad_norm: -1
+  check_val_every_n_epoch: 1
+  fast_dev_run: false
+  accumulate_grad_batches: null
+  max_epochs: null
+  min_epochs: null
+  max_steps: 1000000
+  min_steps: null
+  max_time: null
+  limit_train_batches: null
+  limit_val_batches: 100
+  limit_test_batches: null
+  limit_predict_batches: null
+  val_check_interval: null
+  log_every_n_steps: 50
+  accelerator: gpu
+  strategy: ddp
+  sync_batchnorm: false
+  precision: 32
+  enable_model_summary: true
+  num_sanity_val_steps: 2
+  resume_from_checkpoint: null
+  profiler: null
+  benchmark: null
+  deterministic: null
+  reload_dataloaders_every_n_epochs: 0
+  auto_lr_find: false
+  replace_sampler_ddp: true
+  detect_anomaly: false
+  auto_scale_batch_size: false
+  plugins: null
+  amp_backend: native
+  amp_level: null
+  move_metrics_to_cpu: false
+  multiple_trainloader_mode: max_size_cycle
+  inference_mode: true
+model:
+  class_path: vocos.experiment.MiMoWavLMVAEExp
+  init_args:
+    feature_extractor:
+      class_path: vocos.feature_extractors.WavLMVAEFeatures
+      init_args:
+        model_id: ckpts/wavlm-large
+        layer_idx: -1
+        freeze_model: true
+        latent_dim: 128
+        stage: 2
+        stage1_ckpt_path: logs/wavlmvae-mimo-librispeech-stage1_kl1e-4_ae_300mdeco/first/version_2/checkpoints/vocos_checkpoint_epoch=219_step=138000_val_loss=6.6572.ckpt
+        use_vae: false
+        use_sigma_vae: false
+        use_temporal_downsampling: false
+    backbone:
+      class_path: vocos.models.MiMoBackbone
+      init_args:
+        d_model: 1024
+        decoder_attention_heads: 16
+        decoder_ffn_dim: 4096
+        sampling_rate: 16000
+        hop_length: 160
+        window_size: 640
+        nfft: 640
+        upsample: true
+        latent_dim: 128
+        decoder_layers: 24
+    head: null
+    sample_rate: 16000
+    initial_learning_rate: 0.0001
+    num_warmup_steps: 5000
+    mel_loss_coeff: 4.5
+    mrd_loss_coeff: 1.0
+    kl_loss_coeff: 0.0001
+    sr_loss_coeff: 1.0
+    gan_loss_coeff: 0.1
+    pretrain_mel_steps: 0
+    decay_mel_coeff: false
+    evaluate_utmos: true
+    evaluate_pesq: true
+    evaluate_periodicty: false
+    evaluate_stoi: false
+    evaluate_pesq_wb: false
+    evaluate_sim: true
+data:
+  class_path: vocos.dataset.VocosDataModule
+  init_args:
+    train_params:
+      filelist_path: data/librispeech_train
+      sampling_rate: 16000
+      num_samples: 160000
+      batch_size: 8
+      num_workers: 8
+    val_params:
+      filelist_path: data/librispeech_test_clean
+      sampling_rate: 16000
+      num_samples: 160000
+      batch_size: 8
+      num_workers: 8
+optimizer: null
+lr_scheduler: null

WavCube/hparams.yaml

@@ -0,0 +1,16 @@
+sample_rate: 16000
+initial_learning_rate: 0.0001
+num_warmup_steps: 5000
+mel_loss_coeff: 4.5
+mrd_loss_coeff: 1.0
+pretrain_mel_steps: 0
+decay_mel_coeff: false
+evaluate_utmos: true
+evaluate_pesq: true
+evaluate_periodicty: false
+evaluate_stoi: false
+evaluate_pesq_wb: false
+evaluate_sim: true
+kl_loss_coeff: 0.0001
+sr_loss_coeff: 1.0
+gan_loss_coeff: 0.1