Auto feature extractor (#11097)

* AutoFeatureExtractor

* Init and first tests

* Tests

* Damn you gitignore

* Quality

* Defensive test for when not all backends are here

* Use pattern for Speech2Text models

tests/fixtures/dummy_feature_extractor_config.json

@@ -0,0 +1,3 @@
+{
+  "feature_extractor_type": "Wav2Vec2FeatureExtractor"
+}

tests/test_feature_extraction_auto.py

@@ -0,0 +1,44 @@
+# coding=utf-8
+# Copyright 2021 the HuggingFace Inc. team.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import os
+import unittest
+
+from transformers import FEATURE_EXTRACTOR_MAPPING, AutoFeatureExtractor, Wav2Vec2FeatureExtractor
+
+
+SAMPLE_FEATURE_EXTRACTION_CONFIG = os.path.join(
+    os.path.dirname(os.path.abspath(__file__)), "fixtures/dummy_feature_extractor_config.json"
+)
+
+
+class AutoFeatureExtractorTest(unittest.TestCase):
+    def test_feature_extractor_from_model_shortcut(self):
+        config = AutoFeatureExtractor.from_pretrained("facebook/wav2vec2-base-960h")
+        self.assertIsInstance(config, Wav2Vec2FeatureExtractor)
+
+    def test_feature_extractor_from_local_file(self):
+        config = AutoFeatureExtractor.from_pretrained(SAMPLE_FEATURE_EXTRACTION_CONFIG)
+        self.assertIsInstance(config, Wav2Vec2FeatureExtractor)
+
+    def test_pattern_matching_fallback(self):
+        """
+        In cases where config.json doesn't include a model_type,
+        perform a few safety checks on the config mapping's order.
+        """
+        # no key string should be included in a later key string (typical failure case)
+        keys = list(FEATURE_EXTRACTOR_MAPPING.keys())
+        for i, key in enumerate(keys):
+            self.assertFalse(any(key in later_key for later_key in keys[i + 1 :]))

tests/test_feature_extraction_speech_to_text.py

@@ -20,12 +20,15 @@ import unittest
 
 import numpy as np
 
-from transformers import Speech2TextFeatureExtractor
+from transformers import is_speech_available
 from transformers.testing_utils import require_torch, require_torchaudio
 
 from .test_sequence_feature_extraction_common import SequenceFeatureExtractionTestMixin
 
 
+if is_speech_available():
+    from transformers import Speech2TextFeatureExtractor
+
 global_rng = random.Random()
 
 
@@ -101,7 +104,7 @@ class Speech2TextFeatureExtractionTester(unittest.TestCase):
 @require_torchaudio
 class Speech2TextFeatureExtractionTest(SequenceFeatureExtractionTestMixin, unittest.TestCase):
 
-    feature_extraction_class = Speech2TextFeatureExtractor
+    feature_extraction_class = Speech2TextFeatureExtractor if is_speech_available() else None
 
     def setUp(self):
         self.feat_extract_tester = Speech2TextFeatureExtractionTester(self)

tests/test_processor_speech_to_text.py

@@ -19,7 +19,7 @@ import unittest
 from pathlib import Path
 from shutil import copyfile
 
-from transformers import Speech2TextFeatureExtractor, Speech2TextProcessor, Speech2TextTokenizer
+from transformers import Speech2TextTokenizer, is_speech_available
 from transformers.file_utils import FEATURE_EXTRACTOR_NAME
 from transformers.models.speech_to_text.tokenization_speech_to_text import VOCAB_FILES_NAMES, save_json
 from transformers.testing_utils import require_sentencepiece, require_torch, require_torchaudio
@@ -27,6 +27,10 @@ from transformers.testing_utils import require_sentencepiece, require_torch, req
 from .test_feature_extraction_speech_to_text import floats_list
 
 
+if is_speech_available():
+    from transformers import Speech2TextFeatureExtractor, Speech2TextProcessor
+
+
 SAMPLE_SP = os.path.join(os.path.dirname(os.path.abspath(__file__)), "fixtures/test_sentencepiece.model")