ASR HUGGINGFACE_BACKBONES use AutoModel

CHANGELOG.md

@@ -28,6 +28,8 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 
 - Updated `FlashFinetuning` callback to use separate hooks that lets users use the freezing logic provided out-of-the-box from flash, route FlashFinetuning through a registry. ([#830](https://github.com/PyTorchLightning/lightning-flash/pull/830))
 
+- Changed the `SpeechRecognition` task to use `AutoModelForCTC` rather than just `Wav2Vec2ForCTC` ([#874](https://github.com/PyTorchLightning/lightning-flash/pull/874))
+
 ### Deprecated
 
 - Deprecated `flash.core.data.process.Serializer` in favour of `flash.core.data.io.output.Output` ([#927](https://github.com/PyTorchLightning/lightning-flash/pull/927))

flash/audio/speech_recognition/backbone.py

@@ -20,7 +20,7 @@
 SPEECH_RECOGNITION_BACKBONES = FlashRegistry("backbones")
 
 if _AUDIO_AVAILABLE:
-    from transformers import Wav2Vec2ForCTC
+    from transformers import AutoModelForCTC, Wav2Vec2ForCTC
 
     WAV2VEC_MODELS = ["facebook/wav2vec2-base-960h", "facebook/wav2vec2-large-960h-lv60"]
 
@@ -31,6 +31,6 @@
             providers=[_HUGGINGFACE, _FAIRSEQ],
         )
 
-    HUGGINGFACE_BACKBONES = ExternalRegistry(Wav2Vec2ForCTC.from_pretrained, "backbones", providers=_HUGGINGFACE)
+    HUGGINGFACE_BACKBONES = ExternalRegistry(AutoModelForCTC.from_pretrained, "backbones", providers=_HUGGINGFACE)
 
     SPEECH_RECOGNITION_BACKBONES += HUGGINGFACE_BACKBONES

requirements/datatype_audio.txt

@@ -1,4 +1,4 @@
 torchaudio
 librosa>=0.8.1
-transformers>=4.5
+transformers>=4.11.0
 datasets>=1.8