Dataset for the pretrained model and inquiry about the last layer of the model

Hi,

1. Could you please also share the dataset used to train this model?

2. In [https://github.com/davidsandberg/facenet/wiki/Training-using-the-VGGFace2-dataset](FaceNet github) they decided to use 512 embeddings instead of 128 while I noticed you use 128. Can you also share a "512" similar updated pre-trained model for TF 2.X?