Spaces:
Runtime error
Runtime error
| from transformers import Wav2Vec2Processor, Wav2Vec2ForCTC | |
| import gradio as gr | |
| import torch | |
| import soundfile as sf | |
| import librosa | |
| # load model and processor | |
| model_name = "OthmaneJ/distil-wav2vec2" | |
| processor = Wav2Vec2Processor.from_pretrained(model_name) | |
| model = Wav2Vec2ForCTC.from_pretrained(model_name) | |
| # quantization | |
| # model.eval() | |
| # model_int8 = torch.quantization.quantize_dynamic(model, dtype=torch.qint8,inplace = True,) | |
| # define function to read in sound file | |
| # def map_to_array(file): | |
| # speech, sample_rate = sf.read(file) | |
| # return speech, sample_rate | |
| # tokenize | |
| def inference(audio): | |
| # read in sound file | |
| speech, _ = librosa.load(audio.name,sr=16_000) | |
| input_values = processor(speech, sampling_rate=16_000, return_tensors="pt", padding="longest").input_values # Batch size 1 | |
| # retrieve logits | |
| logits = model(input_values).logits | |
| # take argmax and decode | |
| predicted_ids = torch.argmax(logits, dim=-1) | |
| transcription = processor.batch_decode(predicted_ids) | |
| return transcription[0] | |
| inputs = gr.inputs.Audio(label="Input Audio", type="file") | |
| outputs = gr.outputs.Textbox(label="Output Text") | |
| title = "distilled wav2vec 2.0" | |
| description = "Gradio demo for a distilled wav2vec 2.0 (4x faster than large wav2vec 2.0, and 16x times smaller than base wav2vec 2.0 if combined with quantization). To use it, simply upload your audio, or click one of the examples to load them. Read more at the links below. Currently supports .wav and .flac files" | |
| article = "<p style='text-align: center'><a href='https://github.com/OthmaneJ/distil-wav2vec2' target='_blank'> Github repo for demonstration </a> | <a href='https://huggingface.co/OthmaneJ/distil-wav2vec2' target='_blank'>Pretrained model</a></p>" | |
| examples=[['looked-like-a-stiff-breeze-would-blow-him-over-that-was-my-first-impression-of-the-man.wav']] | |
| gr.Interface(inference, inputs, outputs, title=title, description=description, article=article, examples=examples).launch() |