Training on jetson nano is killed

I’m using Jetson nano 8gb to train the mmdet3d model on customize dataset. However when I started training, it couldn’t last long before getting killed (code’9’) even if i’m using very small size of dataset. I’m not sure but I think it might be because the memory is not enough so I check it using jtop and it shows this before getting killed

Can someone help me? I don’t know where’s the problem or what did I do wrong.

Hi @62990023, yes all your RAM and SWAP is full, so your board ran out of memory. You could try some of these suggestions like mounting more SWAP, disabling ZRAM, and turning off the desktop GUI to save memory:

However understand that Jetson is not normally a training device for DNNs, so you would want to use a small batch size (like 1) and the code may need optimized for low memory overhead. There are some training tutorials that I’ve confirmed to work on Jetson here:

1 Like

Hi. Thank you for the suggestion. I tried reducing the batch size to 1, and now it’s able to train. Thank you.

1 Like