no more dummies for speech processors (#21517)

src/transformers/__init__.py

@@ -391,6 +391,7 @@ _import_structure = {
     "models.speech_to_text": [
         "SPEECH_TO_TEXT_PRETRAINED_CONFIG_ARCHIVE_MAP",
         "Speech2TextConfig",
+        "Speech2TextProcessor",
     ],
     "models.speech_to_text_2": [
         "SPEECH_TO_TEXT_2_PRETRAINED_CONFIG_ARCHIVE_MAP",
@@ -403,6 +404,7 @@ _import_structure = {
         "SPEECHT5_PRETRAINED_HIFIGAN_CONFIG_ARCHIVE_MAP",
         "SpeechT5Config",
         "SpeechT5HifiGanConfig",
+        "SpeechT5Processor",
     ],
     "models.splinter": ["SPLINTER_PRETRAINED_CONFIG_ARCHIVE_MAP", "SplinterConfig", "SplinterTokenizer"],
     "models.squeezebert": ["SQUEEZEBERT_PRETRAINED_CONFIG_ARCHIVE_MAP", "SqueezeBertConfig", "SqueezeBertTokenizer"],
@@ -765,19 +767,6 @@ except OptionalDependencyNotAvailable:
 else:
     _import_structure["models.gpt2"].append("TFGPT2Tokenizer")
 
-try:
-    if not (is_sentencepiece_available() and is_speech_available()):
-        raise OptionalDependencyNotAvailable()
-except OptionalDependencyNotAvailable:
-    from .utils import dummy_sentencepiece_and_speech_objects
-
-    _import_structure["utils.dummy_sentencepiece_and_speech_objects"] = [
-        name for name in dir(dummy_sentencepiece_and_speech_objects) if not name.startswith("_")
-    ]
-else:
-    _import_structure["models.speech_to_text"].append("Speech2TextProcessor")
-    _import_structure["models.speecht5"].append("SpeechT5Processor")
-
 # Vision-specific objects
 try:
     if not is_vision_available():
@@ -3852,7 +3841,11 @@ if TYPE_CHECKING:
     from .models.sew import SEW_PRETRAINED_CONFIG_ARCHIVE_MAP, SEWConfig
     from .models.sew_d import SEW_D_PRETRAINED_CONFIG_ARCHIVE_MAP, SEWDConfig
     from .models.speech_encoder_decoder import SpeechEncoderDecoderConfig
-    from .models.speech_to_text import SPEECH_TO_TEXT_PRETRAINED_CONFIG_ARCHIVE_MAP, Speech2TextConfig
+    from .models.speech_to_text import (
+        SPEECH_TO_TEXT_PRETRAINED_CONFIG_ARCHIVE_MAP,
+        Speech2TextConfig,
+        Speech2TextProcessor,
+    )
     from .models.speech_to_text_2 import (
         SPEECH_TO_TEXT_2_PRETRAINED_CONFIG_ARCHIVE_MAP,
         Speech2Text2Config,
@@ -3864,6 +3857,7 @@ if TYPE_CHECKING:
         SPEECHT5_PRETRAINED_HIFIGAN_CONFIG_ARCHIVE_MAP,
         SpeechT5Config,
         SpeechT5HifiGanConfig,
+        SpeechT5Processor,
     )
     from .models.splinter import SPLINTER_PRETRAINED_CONFIG_ARCHIVE_MAP, SplinterConfig, SplinterTokenizer
     from .models.squeezebert import SQUEEZEBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, SqueezeBertConfig, SqueezeBertTokenizer
@@ -4181,15 +4175,6 @@ if TYPE_CHECKING:
     else:
         from .models.gpt2 import TFGPT2Tokenizer
 
-    try:
-        if not (is_speech_available() and is_sentencepiece_available()):
-            raise OptionalDependencyNotAvailable()
-    except OptionalDependencyNotAvailable:
-        from .utils.dummy_sentencepiece_and_speech_objects import *
-    else:
-        from .models.speech_to_text import Speech2TextProcessor
-        from .models.speecht5 import SpeechT5Processor
-
     try:
         if not is_vision_available():
             raise OptionalDependencyNotAvailable()

src/transformers/models/speech_to_text/__init__.py

@@ -25,6 +25,7 @@ from ...utils import (
 
 _import_structure = {
     "configuration_speech_to_text": ["SPEECH_TO_TEXT_PRETRAINED_CONFIG_ARCHIVE_MAP", "Speech2TextConfig"],
+    "processing_speech_to_text": ["Speech2TextProcessor"],
 }
 
 try:
@@ -43,14 +44,6 @@ except OptionalDependencyNotAvailable:
 else:
     _import_structure["feature_extraction_speech_to_text"] = ["Speech2TextFeatureExtractor"]
 
-try:
-    if not (is_speech_available() and is_sentencepiece_available()):
-        raise OptionalDependencyNotAvailable()
-except OptionalDependencyNotAvailable:
-    pass
-else:
-    _import_structure["processing_speech_to_text"] = ["Speech2TextProcessor"]
-
 try:
     if not is_tf_available():
         raise OptionalDependencyNotAvailable()
@@ -80,6 +73,7 @@ else:
 
 if TYPE_CHECKING:
     from .configuration_speech_to_text import SPEECH_TO_TEXT_PRETRAINED_CONFIG_ARCHIVE_MAP, Speech2TextConfig
+    from .processing_speech_to_text import Speech2TextProcessor
 
     try:
         if not is_sentencepiece_available():
@@ -97,14 +91,6 @@ if TYPE_CHECKING:
     else:
         from .feature_extraction_speech_to_text import Speech2TextFeatureExtractor
 
-    try:
-        if not (is_speech_available() and is_sentencepiece_available()):
-            raise OptionalDependencyNotAvailable()
-    except OptionalDependencyNotAvailable:
-        pass
-    else:
-        from .processing_speech_to_text import Speech2TextProcessor
-
     try:
         if not is_tf_available():
             raise OptionalDependencyNotAvailable()

src/transformers/models/speecht5/__init__.py

@@ -29,6 +29,7 @@ _import_structure = {
         "SpeechT5Config",
         "SpeechT5HifiGanConfig",
     ],
+    "processing_speecht5": ["SpeechT5Processor"],
 }
 
 try:
@@ -47,14 +48,6 @@ except OptionalDependencyNotAvailable:
 else:
     _import_structure["feature_extraction_speecht5"] = ["SpeechT5FeatureExtractor"]
 
-try:
-    if not (is_speech_available() and is_sentencepiece_available()):
-        raise OptionalDependencyNotAvailable()
-except OptionalDependencyNotAvailable:
-    pass
-else:
-    _import_structure["processing_speecht5"] = ["SpeechT5Processor"]
-
 try:
     if not is_torch_available():
         raise OptionalDependencyNotAvailable()
@@ -78,6 +71,7 @@ if TYPE_CHECKING:
         SpeechT5Config,
         SpeechT5HifiGanConfig,
     )
+    from .processing_speecht5 import SpeechT5Processor
 
     try:
         if not is_sentencepiece_available():
@@ -95,14 +89,6 @@ if TYPE_CHECKING:
     else:
         from .feature_extraction_speecht5 import SpeechT5FeatureExtractor
 
-    try:
-        if not (is_speech_available() and is_sentencepiece_available()):
-            raise OptionalDependencyNotAvailable()
-    except OptionalDependencyNotAvailable:
-        pass
-    else:
-        from .processing_speecht5 import SpeechT5Processor
-
     try:
         if not is_torch_available():
             raise OptionalDependencyNotAvailable()

src/transformers/utils/dummy_sentencepiece_and_speech_objects.py

-# This file is autogenerated by the command `make fix-copies`, do not edit.
-from ..utils import DummyObject, requires_backends
-
-
-class Speech2TextProcessor(metaclass=DummyObject):
-    _backends = ["sentencepiece", "speech"]
-
-    def __init__(self, *args, **kwargs):
-        requires_backends(self, ["sentencepiece", "speech"])
-
-
-class SpeechT5Processor(metaclass=DummyObject):
-    _backends = ["sentencepiece", "speech"]
-
-    def __init__(self, *args, **kwargs):
-        requires_backends(self, ["sentencepiece", "speech"])