[Audio Processor] Only pass sr to feat extractor (#20022)

* [Audio Processor] Only pass sr to feat extractor

* move out of if/else

* copy to other processors

src/transformers/models/mctct/processing_mctct.py

@@ -58,6 +58,7 @@ class MCTCTProcessor(ProcessorMixin):
             audio = kwargs.pop("raw_speech")
         else:
             audio = kwargs.pop("audio", None)
+        sampling_rate = kwargs.pop("sampling_rate", None)
         text = kwargs.pop("text", None)
         if len(args) > 0:
             audio = args[0]
@@ -67,7 +68,7 @@ class MCTCTProcessor(ProcessorMixin):
             raise ValueError("You need to specify either an `audio` or `text` input to process.")
 
         if audio is not None:
-            inputs = self.feature_extractor(audio, *args, **kwargs)
+            inputs = self.feature_extractor(audio, *args, sampling_rate=sampling_rate, **kwargs)
         if text is not None:
             encodings = self.tokenizer(text, **kwargs)
 

src/transformers/models/speech_to_text/processing_speech_to_text.py

@@ -61,6 +61,7 @@ class Speech2TextProcessor(ProcessorMixin):
             audio = kwargs.pop("raw_speech")
         else:
             audio = kwargs.pop("audio", None)
+        sampling_rate = kwargs.pop("sampling_rate", None)
         text = kwargs.pop("text", None)
         if len(args) > 0:
             audio = args[0]
@@ -70,7 +71,7 @@ class Speech2TextProcessor(ProcessorMixin):
             raise ValueError("You need to specify either an `audio` or `text` input to process.")
 
         if audio is not None:
-            inputs = self.feature_extractor(audio, *args, **kwargs)
+            inputs = self.feature_extractor(audio, *args, sampling_rate=sampling_rate, **kwargs)
         if text is not None:
             encodings = self.tokenizer(text, **kwargs)
 

src/transformers/models/speech_to_text_2/processing_speech_to_text_2.py

@@ -60,6 +60,7 @@ class Speech2Text2Processor(ProcessorMixin):
             audio = kwargs.pop("raw_speech")
         else:
             audio = kwargs.pop("audio", None)
+        sampling_rate = kwargs.pop("sampling_rate", None)
         text = kwargs.pop("text", None)
         if len(args) > 0:
             audio = args[0]
@@ -69,7 +70,7 @@ class Speech2Text2Processor(ProcessorMixin):
             raise ValueError("You need to specify either an `audio` or `text` input to process.")
 
         if audio is not None:
-            inputs = self.feature_extractor(audio, *args, **kwargs)
+            inputs = self.feature_extractor(audio, *args, sampling_rate=sampling_rate, **kwargs)
         if text is not None:
             encodings = self.tokenizer(text, **kwargs)
 

src/transformers/models/wav2vec2/processing_wav2vec2.py

@@ -80,6 +80,7 @@ class Wav2Vec2Processor(ProcessorMixin):
             audio = kwargs.pop("raw_speech")
         else:
             audio = kwargs.pop("audio", None)
+        sampling_rate = kwargs.pop("sampling_rate", None)
         text = kwargs.pop("text", None)
         if len(args) > 0:
             audio = args[0]
@@ -89,7 +90,7 @@ class Wav2Vec2Processor(ProcessorMixin):
             raise ValueError("You need to specify either an `audio` or `text` input to process.")
 
         if audio is not None:
-            inputs = self.feature_extractor(audio, *args, **kwargs)
+            inputs = self.feature_extractor(audio, *args, sampling_rate=sampling_rate, **kwargs)
         if text is not None:
             encodings = self.tokenizer(text, **kwargs)
 

src/transformers/models/wav2vec2_with_lm/processing_wav2vec2_with_lm.py

@@ -228,6 +228,7 @@ class Wav2Vec2ProcessorWithLM(ProcessorMixin):
             audio = kwargs.pop("raw_speech")
         else:
             audio = kwargs.pop("audio", None)
+        sampling_rate = kwargs.pop("sampling_rate", None)
         text = kwargs.pop("text", None)
         if len(args) > 0:
             audio = args[0]
@@ -237,7 +238,7 @@ class Wav2Vec2ProcessorWithLM(ProcessorMixin):
             raise ValueError("You need to specify either an `audio` or `text` input to process.")
 
         if audio is not None:
-            inputs = self.feature_extractor(audio, *args, **kwargs)
+            inputs = self.feature_extractor(audio, *args, sampling_rate=sampling_rate, **kwargs)
         if text is not None:
             encodings = self.tokenizer(text, **kwargs)
 

src/transformers/models/whisper/processing_whisper.py

@@ -85,6 +85,7 @@ class WhisperProcessor(ProcessorMixin):
             return self.current_processor(*args, **kwargs)
 
         audio = kwargs.pop("audio", None)
+        sampling_rate = kwargs.pop("sampling_rate", None)
         text = kwargs.pop("text", None)
         if len(args) > 0:
             audio = args[0]
@@ -94,7 +95,7 @@ class WhisperProcessor(ProcessorMixin):
             raise ValueError("You need to specify either an `audio` or `text` input to process.")
 
         if audio is not None:
-            inputs = self.feature_extractor(audio, *args, **kwargs)
+            inputs = self.feature_extractor(audio, *args, sampling_rate=sampling_rate, **kwargs)
         if text is not None:
             encodings = self.tokenizer(text, **kwargs)