Can I load 3 different models into a single graphic card?

[ejentos]

Hi everyone,
Currently I use CPU to process audio files with the code below. All loaded models take ~11.5 GB RAM.

    model_names = ["small.en", "medium.en", "large"]
    model_map = {name: whisper.load_model(name) for name in model_names}

    while True:
        try:
            records = get_records(url)
            for record in records:
                 ......
                model = model_map[model_name]
                transcription = model.transcribe(file, language='en', task='transcribe', word_timestamps=True, initial_prompt=config["Whisper"]["prompt"])

I wonder:

1. if I can use the same code to load 3 models into a graphic card?
2. Would it be sufficient to have a graphic card with 16 GB vRAM? According to the 'Required VRAM' from the table it should consume ~2 + 5 + 10 = 17 vRAM, but if the case is the same when I load 3 models together?

Best Regards