how to load model and restore training tensorflow
I would recommend saving the whole model with model.save(*) and then loading it again with model.load(*). See this documentation for more information. In your case, you can just run:
model.load_weights('teta/your_checkpoint')
before calling model.fit(*) again.