| # import | |
| import librosa, torch | |
| from transformers import Wav2Vec2ForCTC, Wav2Vec2Tokenizer | |
| # load the tokenizer and model | |
| tokenizer = Wav2Vec2Tokenizer.from_pretrained("facebook/wav2vec2-large-960h") | |
| model = Wav2Vec2ForCTC.from_pretrained("facebook/wav2vec2-large-960h") | |
| # load the audio data (use your own wav file here!) | |
| input_audio, sr = librosa.load('my_wav_file.wav', sr=16000) | |
| # tokenize | |
| input_values = tokenizer(input_audio, return_tensors="pt", padding="longest").input_values | |
| # retrieve logits | |
| logits = model(input_values).logits | |
| # take argmax and decode | |
| transcription = tokenizer.batch_decode(torch.argmax(logits, dim=-1)) | |
| # print the output | |
| print(transcription) | |