Upload tokaiteio.json

configs/tokaiteio.json

@@ -0,0 +1,53 @@
+{
+  "train": {
+    "log_interval": 50,
+    "eval_interval": 500,
+    "seed": 1234,
+    "epochs": 20000,
+    "learning_rate": 2e-4,
+    "betas": [0.8, 0.99],
+    "eps": 1e-9,
+    "batch_size": 16,
+    "fp16_run": true,
+    "lr_decay": 0.999875,
+    "segment_size": 8192,
+    "init_lr_ratio": 1,
+    "warmup_epochs": 0,
+    "c_mel": 45,
+    "c_kl": 1.0
+  },
+  "data": {
+    "training_files":"filelists/train.txt.cleaned",
+    "validation_files":"filelists/val.txt.cleaned",
+    "text_cleaners":["japanese_phrase_cleaners"],
+    "max_wav_value": 32768,
+    "sampling_rate": 22050,
+    "filter_length": 1024,
+    "hop_length": 256,
+    "win_length": 1024,
+    "n_mel_channels": 80,
+    "mel_fmin": 0.0,
+    "mel_fmax": null,
+    "add_blank": true,
+    "n_speakers": 0,
+    "cleaned_text": true
+  },
+  "model": {
+    "inter_channels": 192,
+    "hidden_channels": 192,
+    "filter_channels": 768,
+    "n_heads": 2,
+    "n_layers": 6,
+    "kernel_size": 3,
+    "p_dropout": 0.1,
+    "resblock": "1",
+    "resblock_kernel_sizes": [3,7,11],
+    "resblock_dilation_sizes": [[1,3,5], [1,3,5], [1,3,5]],
+    "upsample_rates": [8,8,2,2],
+    "upsample_initial_channel": 512,
+    "upsample_kernel_sizes": [16,16,4,4],
+    "n_layers_q": 3,
+    "use_spectral_norm": false
+  },
+    "model_dir": "../autodl-tmp/logs/tokaiteio_model"
+}