wav2vec2-large-xlsr-53-english caikit demo

This demos sends the same audio to the HuggingFace automatic-speech-recognition pipeline default model facebook/wav2vec2-base-960h and the jonatasgrosman/wav2vec2-large-xlsr-53-english model.

It uses the caikit to serve the model.

Before Starting

The following tools are required:

• python (v3.8+)
• pip (v23.0+)

Note: Before installing dependencies and to avoid conflicts in your environment, it is advisable to use a virtual environment(venv).

Install the dependencies: pip install -r requirements.txt

Running the Caikit runtime

In one terminal, start the runtime server:

python app.py

Wait it downlaod the models and start the gradio server:

Running on local URL:  http://127.0.0.1:8080

Access the browser, select a sample audio or record you own to test it out!