Google Colab

You are familiar with the intense computational requirements of deep learning models. On a CPU, it would take a remarkably long time to train a deep learning model with lots of training data. Hence, to keep training times practical, it is common practice to use cloud-based services that offer Graphics Processing Units (GPU) to speed up computations. You can expect a speedup of 10-30 times when compared to running the training session on a CPU. The exact amount of speedup, of course, depends upon the power of the GPU, the amount of data involved, and the processing steps.

There are many vendors offering such cloud services, such as Amazon Web Services (AWS), Microsoft Azure and others. Google offers an environment/IDE called Google Colab, which offers up to 12 hours of free GPU usage per day for anyone looking to train deep learning models. Additionally, the code is run on a Jupyter-like notebook. In this chapter, we will leverage the power of Google Colab to develop our deep...