Title: Whisper (Medium) Model Getting “Killed” on Jetson Orin Nano

Robotics & Edge Computing Jetson & Embedded Systems Jetson Orin Nano
1

i’m trying to run OpenAI’s Whisper medium model on a Jetson Orin Nano, but the process always dies with an OOM (“killed”) error. So far, I’ve allocated over 64 GB of swap, set vm.swappiness to 80, made sure there’s plenty of free disk space, and I’m running with CUDA enabled—but nothing seems to help.

This is my first time posting on the NVIDIA forum, so if I’ve formatted this poorly or missed any important details, please let me know. Any suggestions or troubleshooting tips would be hugely appreciated!

Thanks in advance.

3

Hi,

Since Orin Nano has limited memory, could you try a lightweight model (<= 4B) to see if it can work?

Below is our tutorial for Whisper for your reference:

It should work on Orin Nano 8GB, but please run the memory optimization steps shared in the link below first:

Thanks.