fix bug

paddlespeech/s2t/exps/whisper/test_wav.py

@@ -45,6 +45,7 @@ class WhisperInfer():
         model_dict = paddle.load(self.config.model_file)
         config.pop("model_file")
         dims = ModelDimensions(**model_dict["dims"])
+        self.dims = dims
         self.model = Whisper(dims)
         self.model.load_dict(model_dict)
 
@@ -63,12 +64,10 @@ class WhisperInfer():
                 temperature = [temperature]
 
             #load audio
-            # mel = log_mel_spectrogram(
+            mel = log_mel_spectrogram(
-            #     args.audio_file, resource_path=config.resource_path, , n_mels=128)
-            audio = log_mel_spectrogram(
                 args.audio_file,
                 resource_path=config.resource_path,
-                n_mels=128,
+                n_mels=self.dims.n_mels,
                 padding=480000)
             result = transcribe(
                 self.model, mel, temperature=temperature, **config)

paddlespeech/s2t/models/whisper/whisper.py

@@ -397,9 +397,7 @@ def detect_language(
 
     # skip encoder forward pass if already-encoded audio features were given
     if mel.shape[-2:] != (model.dims.n_audio_ctx, model.dims.n_audio_state):
-        mel = model.encoder(
+        mel = model.encoder(mel)
-            mel
-        )  # TODO zhaoxi: torch return float16, while cause e-3 diff with paddle float32
 
     # forward pass using a single token, startoftranscript
     batch_size = mel.shape[0]
@@ -1149,7 +1147,6 @@ class DecodingTask:
         self.options: DecodingOptions = self._verify_options(options)
         self.resource_path: str = resource_path
 
-        # self.beam_size: int = options.beam_size or options.best_of or 1
         self.n_group: int = options.beam_size or options.best_of or 1
         self.n_ctx: int = model.dims.n_text_ctx
         self.sample_len: int = options.sample_len or model.dims.n_text_ctx // 2