<< return to Vizycam.com

New model training in google never finishes

Hi. Trying to add a new object model in Object Detector. I followed all the steps, but can’t complete the training in google. It just stays in the first step crunching until the disk goes from 28 GB to 107 GB and errors out when the disk is full. This first step runs for 15+ mins before the disk is full and then errors. Feels like I’m missing something here. Tried all my runtime types to see if it makes a difference.



Well doing some digging it’s the same problem as the other thread about Google Colab. It seems like TFlite model maker has been broken since May.

So what’s the alternative? The cam is not useful to me if I can make new models.

Can you switch over to Mediapipe Tensorflow lite model? I don’t know when the current one will be fixed. It’s been broken since April/May. :frowning:

Sorry, we are working on fixing. :frowning:


I have a solution from the tensor forum that I just verified works and made a new model if you’re interested.

Absolutely – please share :slight_smile: