This repo contains code for finetuning wav2vec models on datasets in the Mozilla common voices dataset.
The finetuning script can be run on Google Colab.
The code was inpired by Fine-Tune Wav2Vec Bert for low resource ASR with transformers
The model is Wav2Vec Bert model with a WER of 19.33.
You can download the model from Huggingface and use it directly to produce the same results.