Tensorflow data pipeline improvements

I currently have a google colab file with a data generator and a model. I train the model using the generator. The current training is very slow and I would like to hopefully speed up the training process by using the [login to view URL] api or anything else. Let me know if you are interested in helping. Relevant Files/Documents are attached below. COLAB: https://colab.research.google.com/drive/1oOnjeCP0VuGJOljlJt9fW7EUCCjcmrNp#scrollTo=Ie8-s2T9I7H1 LINK TO JSON DATA: [login to view URL] The JSON data we read is too large to be read into memory all at once, so our preprocessor breaks it up into these smaller json files. I need the dataset to progressively load them rather than all at once.