use ckpt instead of tensors

app.py

@@ -23,8 +23,9 @@ model = load_model(
     DiT,
     dict(dim=1024, depth=22, heads=16, ff_mult=2, text_dim=512, conv_layers=4),
     ckpt_path=str(
-        cached_path("hf://whatvn/vietnamese-tts/model.tensors")
+        cached_path("hf://whatvn/vietnamese-tts/model_960000.pt")
     ),
+    mel_spec_type="vocos",
     vocab_file=str(cached_path("hf://whatvn/vietnamese-tts/vocab.txt")),
 )
 

requirements.txt

@@ -1,4 +1,3 @@
-accelerate
 torch
 torchaudio
 soundfile