初始化项目，由ModelHub XC社区提供模型

Model: iic/InspireMusic-1.5B-Long
Source: Original Platform

wavtokenizer/config.yaml

@@ -0,0 +1,164 @@
+# pytorch_lightning==1.8.6
+seed_everything: 3407
+trainer:
+  logger:
+    class_path: pytorch_lightning.loggers.TensorBoardLogger
+    init_args:
+      save_dir: ./result/
+      name: lightning_logs
+      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: wavtokenizer_checkpoint_{epoch}_{step}_{val_loss:.4f}
+      monitor: val_loss
+      verbose: false
+      save_last: true
+      save_top_k: 10
+      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: inspiremusic.wavtokenizer.decoder.helpers.GradNormCallback
+  default_root_dir: null
+  gradient_clip_val: null
+  gradient_clip_algorithm: null
+  num_nodes: 1
+  num_processes: null
+  devices: -1
+  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: 20000000
+  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: 1000
+  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
+ckpt_path: null
+data:
+  class_path: inspiremusic.wavtokenizer.decoder.dataset.VocosDataModule
+  init_args:
+    train_params:
+      filelist_path: train.scp
+      sampling_rate: 24000
+      num_samples: 72000
+      batch_size: 38
+      num_workers: 8
+    val_params:
+      filelist_path: test.scp
+      sampling_rate: 24000
+      num_samples: 72000
+      batch_size: 10
+      num_workers: 8
+model:
+  class_path: inspiremusic.wavtokenizer.decoder.experiment.WavTokenizer
+  init_args:
+    feature_extractor:
+      class_path: inspiremusic.wavtokenizer.decoder.feature_extractors.EncodecFeatures
+      init_args:
+        encodec_model: encodec_24khz
+        bandwidths:
+        - 6.6
+        - 6.6
+        - 6.6
+        - 6.6
+        train_codebooks: true
+        num_quantizers: 1
+        dowmsamples:
+        - 8
+        - 5
+        - 4
+        - 2
+        vq_bins: 4096
+        vq_kmeans: 200
+    backbone:
+      class_path: inspiremusic.wavtokenizer.decoder.models.VocosBackbone
+      init_args:
+        input_channels: 512
+        dim: 768
+        intermediate_dim: 2304
+        num_layers: 12
+        layer_scale_init_value: null
+        adanorm_num_embeddings: 4
+    head:
+      class_path: inspiremusic.wavtokenizer.decoder.heads.ISTFTHead
+      init_args:
+        dim: 768
+        n_fft: 1280
+        hop_length: 320
+        padding: same
+    resume_config: config.yaml
+    resume_model: last.ckpt
+    sample_rate: 24000
+    initial_learning_rate: 0.0001
+    num_warmup_steps: 0
+    mel_loss_coeff: 45.0
+    mrd_loss_coeff: 1.0
+    pretrain_mel_steps: 0
+    decay_mel_coeff: false
+    evaluate_utmos: false
+    evaluate_pesq: true
+    evaluate_periodicty: true
+    resume: true
+
+
+