I am testing a Whisper V3 speech-to-text model from Hugging Face in my Google Colab free version and there are 12,7GB of RAM and 29GB used from 107GB storage.<p>It keeps crashing with "insufficient RAM" error, even after I have uploaded a small file for testing.<p>Why does it keep crushing instead of using more time to complete?
Why doesn't it use/create more swap memory on the free 107GB storage?<p>I just want to test the model to see if it works at all, and then if necessary proceed to more RAM.<p>On my laptop I have ~16GB of RAM. Will it suffice to run the model?<p>If I test this model on AWS free tier, will there be enough RAM etc for initial testing?
I am testing this model <a href="https://huggingface.co/openai/whisper-large-v3" rel="nofollow">https://huggingface.co/openai/whisper-large-v3</a>, and I installed accelerate.