Modify a Stacked Denoising LSTM Autoencoder

I have a database with 90 million lines. That database is essentially features for ML purpose. It is a three years database with one-second frequency on DCIX stock (NASDAQ). I noticed that most of the row features are zeroes. So I want to take each day and verify if the standard deviation on the midprice is greater than 0.05 before passing it to the autoencoder. So your task is to insert the logic of [login to view URL] into my stack denoising LSTM autoencoder. I will guide you before starting the task, and show you the most important things to know about the task remotely on my server.