Memory lack! with Tensorflow oom on GPU about facerecognition

[habrman]

Hi,

First of all, sorry for such a late response but I think you misunderstand the purpose of this repo. I never learn anything. It's more of a showcase repo to show what the model is capable of and I only use a pretrained model to run inference. Please check the readme for links to the code for training. However, reducing the batch size is a good solution to make this code work with a large number of identities. I'm not sure about if you manage to do this by yourself or if you need help to do it?