Update meldataset.py

meldataset.py

@@ -97,6 +97,7 @@ def preprocess(wave: np.ndarray, sample_rate: int = 24000):
     mel = (torch.log(mel + 1e-5) - mean) / std
     return mel.unsqueeze(0)  # (1, n_mels, T)
 
+
 class TextCleaner:
     def __init__(self, symbol_dict, debug=True):
         self.symbol_dict = symbol_dict
@@ -105,15 +106,16 @@ class TextCleaner:
     def __call__(self, text: str):
         text = (text or "").strip()
 
-        tokens = list(text)   # ✅ luôn char-level
+        # ✅ LUÔN char-level như lúc build dataset
+        tokens = list(text)
 
         ids = []
         missing = []
-        for t in tokens:
-            if t in self.symbol_dict:
-                ids.append(self.symbol_dict[t])
+        for ch in tokens:
+            if ch in self.symbol_dict:
+                ids.append(self.symbol_dict[ch])
             else:
-                missing.append(t)
+                missing.append(ch)
 
         if self.debug and missing:
             print(f"[TextCleaner] missing {len(missing)} symbols. sample={missing[:30]}")
@@ -122,6 +124,7 @@ class TextCleaner:
 
 
 
+
 class FilePathDataset(torch.utils.data.Dataset):
     def __init__(
         self,