Training Option : Hardware Acceleration

This section introduces the different hardware options available for training.

Hardware acceleration has two categories of choices, one being the GPU type and then the other being the number of GPUs. Simply put, the more advanced and powerful the GPU is, as well as the number of GPUs used, the more computing power and speed can be given to you.

Why do I need hardware acceleration?

This can affect your training in many different ways. For example, less computing power can restrict the batch size you select in Training Option : Advanced Evaluation. If time to train your model is a limiting factor for you, the difference in time used for training on our platform can be very significant. Faster training times will enable you to have more time to evaluate model performance, tweak and improve model performance, or try other workflows.

How is the usage of computing resources measured?

The way we evaluate computing resource usage is through Compute Minutes. The computation of Compute Minutes can be found in Usage Quota. To see how much of your Compute Minutes has been used, go to Usage Quota. If you want to see other plans that match your amount of Compute Minutes needed, go to Plans and Pricing.

GPU TypeNumber of GPUsDescription
Nvidia T41, 2, 4, 8This is the baseline GPU offered.
Nvidia K801, 2, 4, 8The K80 is a more modern, advanced version of the T4.
Nvidia P1001, 2, 4, 8The P100 has more capacity and faster processing speed than the first two.
Nvidia V1001, 2, 4, 8The V100 is the most advanced data center GPU to be used for AI.