Add files via upload

configs/config.json

@@ -1,103 +1,62 @@
 {
   "train": {
-    "log_interval": 50,
-    "eval_interval": 1000,
+    "log_interval": 200,
+    "eval_interval": 800,
     "seed": 1234,
-    "port": 8001,
     "epochs": 10000,
-    "learning_rate": 0.0002,
+    "learning_rate": 0.0001,
     "betas": [
       0.8,
       0.99
     ],
     "eps": 1e-09,
     "batch_size": 6,
-    "accumulation_steps": 1,
     "fp16_run": false,
-    "lr_decay": 0.998,
+    "lr_decay": 0.999875,
     "segment_size": 10240,
     "init_lr_ratio": 1,
     "warmup_epochs": 0,
     "c_mel": 45,
-    "keep_ckpts":4
+    "c_kl": 1.0,
+    "use_sr": true,
+    "max_speclen": 512,
+    "port": "8001",
+    "keep_ckpts": 3
   },
   "data": {
-    "data_dir": "dataset",
-    "dataset_type": "SingDataset",
-    "collate_type": "SingCollate",
-    "training_filelist": "filelists/train.txt",
-    "validation_filelist": "filelists/val.txt",
+    "training_files": "filelists/train.txt",
+    "validation_files": "filelists/val.txt",
     "max_wav_value": 32768.0,
     "sampling_rate": 44100,
-    "n_fft": 2048,
-    "fmin": 0,
-    "fmax": 22050,
+    "filter_length": 2048,
     "hop_length": 512,
-    "win_size": 2048,
-    "acoustic_dim": 80,
-    "c_dim": 256,
-    "min_level_db": -115,
-    "ref_level_db": 20,
-    "min_db": -115,
-    "max_abs_value": 4.0,
-    "n_speakers": 200
+    "win_length": 2048,
+    "n_mel_channels": 80,
+    "mel_fmin": 0.0,
+    "mel_fmax": 22050
   },
   "model": {
+    "inter_channels": 192,
     "hidden_channels": 192,
-    "spk_channels": 192,
     "filter_channels": 768,
     "n_heads": 2,
-    "n_layers": 4,
+    "n_layers": 6,
     "kernel_size": 3,
     "p_dropout": 0.1,
-    "prior_hidden_channels": 192,
-    "prior_filter_channels": 768,
-    "prior_n_heads": 2,
-    "prior_n_layers": 4,
-    "prior_kernel_size": 3,
-    "prior_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, 2],
+    "upsample_initial_channel": 512,
+    "upsample_kernel_sizes": [16,16, 4, 4, 4],
+    "n_layers_q": 3,
     "use_spectral_norm": false,
-    "resblock_kernel_sizes": [
-      3,
-      7,
-      11
-    ],
-    "resblock_dilation_sizes": [
-      [
-        1,
-        3,
-        5
-      ],
-      [
-        1,
-        3,
-        5
-      ],
-      [
-        1,
-        3,
-        5
-      ]
-    ],
-    "upsample_rates": [
-      8,
-      8,
-      4,
-      2
-    ],
-    "upsample_initial_channel": 256,
-    "upsample_kernel_sizes": [
-      16,
-      16,
-      8,
-      4
-    ],
-    "n_harmonic": 64,
-    "n_bands": 65
+    "gin_channels": 256,
+    "ssl_dim": 256,
+    "n_speakers": 200
   },
   "spk": {
-    "jishuang": 0,
+    "nyaru": 0,
     "huiyu": 1,
     "nen": 2,
     "paimon": 3,

inference/infer_tool.py

@@ -127,8 +127,9 @@ def __init__(self, net_g_path, config_path,
     def load_model(self):
         # 获取模型配置
         self.net_g_ms = SynthesizerTrn(
-            self.hps_ms
-        )
+            self.hps_ms.data.filter_length // 2 + 1,
+            self.hps_ms.train.segment_size // self.hps_ms.data.hop_length,
+            **self.hps_ms.model)
         _ = utils.load_checkpoint(self.net_g_path, self.net_g_ms, None)
         if "half" in self.net_g_path and torch.cuda.is_available():
             _ = self.net_g_ms.half().eval().to(self.dev)
@@ -173,7 +174,7 @@ def infer(self, speaker, tran, raw_path,
             c = c.half()
         with torch.no_grad():
             start = time.time()
-            audio = self.net_g_ms.infer(c, f0=f0, g=sid, uv=uv, predict_f0=auto_predict_f0, noice_scale=noice_scale)[0][0,0].data.float()
+            audio = self.net_g_ms.infer(c, f0=f0, g=sid, uv=uv, predict_f0=auto_predict_f0, noice_scale=noice_scale)[0,0].data.float()
             use_time = time.time() - start
             print("vits use time:{}".format(use_time))
         return audio, audio.shape[-1]