fix missing import (#15016)

Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

src/transformers/models/wav2vec2/modeling_wav2vec2.py

@@ -1606,6 +1606,7 @@ class Wav2Vec2ForMaskedLM(Wav2Vec2PreTrainedModel):
         >>> from transformers import Wav2Vec2Processor, Wav2Vec2ForMaskedLM
         >>> from datasets import load_dataset
         >>> import soundfile as sf
+        >>> import torch
 
         >>> processor = Wav2Vec2Processor.from_pretrained("facebook/wav2vec2-base-960h")
         >>> model = Wav2Vec2ForMaskedLM.from_pretrained("facebook/wav2vec2-base-960h")