use newer model

inference.py

@@ -18,7 +18,7 @@ SampleFn = Callable[
 ]
 ### Loading
 
-def load_model_from_hf(repo_id: str = 'jordand/echo-tts', device: str = 'cuda', dtype: torch.dtype | None = torch.bfloat16, compile: bool = False, token: str | None = None) -> EchoDiT:
+def load_model_from_hf(repo_id: str = 'jordand/echo-tts-base', device: str = 'cuda', dtype: torch.dtype | None = torch.bfloat16, compile: bool = False, token: str | None = None) -> EchoDiT:
     with torch.device('meta'):
         model = EchoDiT(
             latent_size=80, model_size=2048, num_layers=24, num_heads=16,