I am working with image classification and Tensorflow on Google Cloud Machine Learning. Everytime I train a new project, It works perfectly, but everytime I have an update on my dataset (upload new images to database) I have to retrain everything. Is there a way to retrain from the last checkpoint, how should I do that in the cloud?
Thanks everyone!
Tensorflow will continue training from the previous checkpoint. Make sure to specify a cloud storage location as the output directory when you create the model.