Retraining SSD-Mobilenet Issue

I am working through Dusty’s Retraining SSD Mobilenet right now after completing all the tutorials in the Jetson Inference Repo up to this point. I am trying to download my own data sets from Open Images and am getting a traceback error to the python code downloaded from the repo named “open_images_downloader.” The error I am getting is shown below and I don’t understand by this would not be working if the commands and code that the Jetson is using is all from the repo? Any help would be appreciated.

It would appear that there isn’t enough memory to complete the operations - can you keep an eye on the memory usage with sudo tegrastats?

You can also follow these steps to reduce the memory usage of the base OS:

Alright I ran the “sudo init 3” command in my home directory and the monitor screen went black with a white line blinking in the top left corner however it wont let me type any commands of any sort. Do I have to wait for it to reboot?

My task manager also says I am only using 43% of the memory, is this enough to be an issue?

Try pressing Ctrl+Alt+F1 (or similar number keys, I can’t remember exactly which number) and it will give you a shell login.

Is that while the process is running? You might also want to try deleting data/Krill/train-annotations-bbox.csv if it’s corrupted. Normally this should not be much of an issue just to load the bbox annotations.

I utilized the process to obtain 4Gb of swap space and it seemed to work, while running the download the memory definitely filled up and the swap space filled to 50%. Thanks for the help! I am assuming I will need this swap space for almost any training I do on the Jetson considering how much memory it takes?

Yes that is correct (certainly on Nano), and if you need more memory you can temporarily disable the desktop and run the training from terminal or SSH.

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.