[fix] model load dtype

pipeline_ace_step.py

@@ -143,7 +143,7 @@ class ACEStepPipeline:
         self.music_dcae = MusicDCAE(dcae_checkpoint_path=dcae_checkpoint_path, vocoder_checkpoint_path=vocoder_checkpoint_path)
         self.music_dcae.to(device).eval().to(self.dtype)
 
-        self.ace_step_transformer = ACEStepTransformer2DModel.from_pretrained(ace_step_checkpoint_path)
+        self.ace_step_transformer = ACEStepTransformer2DModel.from_pretrained(ace_step_checkpoint_path, torch_dtype=self.dtype)
         self.ace_step_transformer.to(device).eval().to(self.dtype)
 
         lang_segment = LangSegment()
@@ -158,7 +158,7 @@ class ACEStepPipeline:
         ])
         self.lang_segment = lang_segment
         self.lyric_tokenizer = VoiceBpeTokenizer()
-        text_encoder_model = UMT5EncoderModel.from_pretrained(text_encoder_checkpoint_path).eval()
+        text_encoder_model = UMT5EncoderModel.from_pretrained(text_encoder_checkpoint_path, torch_dtype=self.dtype).eval()
         text_encoder_model = text_encoder_model.to(device).to(self.dtype)
         text_encoder_model.requires_grad_(False)
         self.text_encoder_model = text_encoder_model