facebook/data2vec-audio-base-10m

1---2language: en3datasets:4- librispeech_asr5tags:6- speech7 8license: apache-2.09---10 11# Data2Vec-Audio-Base-10m12 13[Facebook's Data2Vec](https://ai.facebook.com/research/data2vec-a-general-framework-for-self-supervised-learning-in-speech-vision-and-language/)14 15The base model pretrained and fine-tuned on 10 minutes of Librispeech on 16kHz sampled speech audio. When using the model16make sure that your speech input is also sampled at 16Khz.17 18[Paper](https://arxiv.org/abs/2202.03555)19 20Authors: Alexei Baevski, Wei-Ning Hsu, Qiantong Xu, Arun Babu, Jiatao Gu, Michael Auli21 22**Abstract**23 24While the general idea of self-supervised learning is identical across modalities, the actual algorithms and objectives differ widely because they were developed with a single modality in mind. To get us closer to general self-supervised learning, we present data2vec, a framework that uses the same learning method for either speech, NLP or computer vision. The core idea is to predict latent representations of the full input data based on a masked view of the input in a self-distillation setup using a standard Transformer architecture. Instead of predicting modality-specific targets such as words, visual tokens or units of human speech which are local in nature, data2vec predicts contextualized latent representations that contain information from the entire input. Experiments on the major benchmarks of speech recognition, image classification, and natural language understanding demonstrate a new state of the art or competitive performance to predominant approaches.25 26The original model can be found under https://github.com/pytorch/fairseq/tree/main/examples/data2vec .27 28# Pre-Training method29 30![model image](https://raw.githubusercontent.com/patrickvonplaten/scientific_images/master/data2vec.png)31 32For more information, please take a look at the [official paper](https://arxiv.org/abs/2202.03555).33 34# Usage35 36To transcribe audio files the model can be used as a standalone acoustic model as follows:37 38```python39 from transformers import Wav2Vec2Processor, Data2VecForCTC40 from datasets import load_dataset41 import torch42 43 # load model and processor44 processor = Wav2Vec2Processor.from_pretrained("facebook/data2vec-audio-base-10m")45 model = Data2VecForCTC.from_pretrained("facebook/data2vec-audio-base-10m")46     47 # load dummy dataset and read soundfiles48 ds = load_dataset("patrickvonplaten/librispeech_asr_dummy", "clean", split="validation")49 50 # tokenize51 input_values = processor(ds[0]["audio"]["array"],, return_tensors="pt", padding="longest").input_values  # Batch size 152 53 # retrieve logits54 logits = model(input_values).logits55 56 # take argmax and decode57 predicted_ids = torch.argmax(logits, dim=-1)58 transcription = processor.batch_decode(predicted_ids)59 ```60