Concepts / Running Deep-Learning Code in a Cloud Environment

Running Deep-Learning Code in a Cloud Environment

A recent NVIDIA GPU is recommended for faster execution, not required for running the examples.

  • Programming

Start with the Execution Location

Before running deep-learning code examples, decide where the computation will take place. The two broad routes described here are a local machine with a recent NVIDIA GPU or a cloud environment that provides access to GPU-based computing.

A recent NVIDIA GPU is recommended because it allows the examples to run faster. However, the recommendation is not a strict requirement: the examples can still be run without that GPU, although execution may be slower.

YesNoRun examplesRecent NVIDIA GPUavailableLocal machineRecommended local setupExecute examplesCloud environmentAlternative route
How do I decide whether to run the examples on a local workstation or use a cloud-based environment?

Recommended Does Not Mean Required

The phrase recommended GPU carries two separate ideas. First, a recent NVIDIA GPU is the preferred local setup because it provides a better-performing experience. Second, it is not a condition that must be satisfied before the examples can run. The GPU changes the expected execution speed, not whether running the examples is permitted by the guidance.

QuestionRecent NVIDIA GPUNo recent NVIDIA GPU
Can the examples be run?YesYes; the GPU is not mandatory
Expected execution speedFaster execution is recommendedThe examples may run more slowly
Role in the setupRecommended local setupUse another available route, including a cloud environment
recommendation, not requirementRecent NVIDIA GPUFaster executionNo recent NVIDIA GPUExamples can still run
What changes when a recent NVIDIA GPU is recommended but not required?

Why GPU Choice Affects Speed

The practical reason to prefer a recent NVIDIA GPU is execution speed. The examples are deep-learning code, and the source specifically recommends this kind of GPU for a faster experience. The important comparison is therefore not whether the code is allowed to run, but how quickly the execution proceeds.

Choosing Between Two Local Setups

A learner can run the examples on a local machine without a recent NVIDIA GPU or on a local machine with a recent NVIDIA GPU. Which setup matches the recommendation?

Identify the recommendation: The source recommends a recent NVIDIA GPU for faster execution.

Separate speed from possibility: The absence of the recommended GPU does not make running the examples impossible because the GPU is not required.

Choose the preferred local route: When both local choices are available, the machine with the recent NVIDIA GPU matches the recommended local setup because it is expected to provide faster execution.

Choose the local machine with the recent NVIDIA GPU when the goal is the recommended, better-performing local experience.

use recent NVIDIA GPUExamplesSlower executionExamplesFaster execution
How does using a recent NVIDIA GPU affect the execution speed of the deep-learning examples?

Cloud Routes Without a Local Workstation

A suitable local workstation is not the only route. Cloud environments can replace the need for local access to a suitable workstation by providing access to GPU-based computing. The two named possibilities in the source are Google Cloud instances and Amazon Web Services GPU instances.

Execution routeWhen it fitsWhat the source identifies
Local machineA recent NVIDIA GPU is available locallyRecommended local setup for a better-performing experience
Google CloudA suitable local workstation is unavailableCloud instance option
Amazon Web ServicesA suitable local workstation is unavailableGPU instance option
alternativealternativeSuitable localworkstationNot availableGoogle Cloud instanceCloud routeAWS GPU instanceCloud route
What execution options are available when I do not have a suitable local workstation?

Make the execution-location decision before working through the examples. If the recommended local setup is available, use it for the better-performing experience. If it is not available, select one of the named cloud routes rather than treating the missing local workstation as a blocker.

Mistakes Beginners Make

  • Treating a recent NVIDIA GPU as mandatory

    The source describes the GPU as recommended for faster execution, not required for running the examples.

    Fix: Recognize that the examples can run without the recommended GPU, although execution may be slower.

  • Assuming that only a local workstation is possible

    Cloud environments can replace the need for access to a suitable local workstation.

    Fix: Consider a Google Cloud instance or an AWS GPU instance.

  • Ignoring the reason for the recommendation

    The stated reason for the recommendation is faster execution and a better-performing experience.

    Fix: Use the recent NVIDIA GPU when available because it is the recommended faster local setup.

Choose Your Route

EASY

You want to run the deep-learning examples. You have no local workstation with a recent NVIDIA GPU, but you can use either a Google Cloud instance or an AWS GPU instance. Which broad execution route should you choose, and why?

Hints
  • First decide whether the recommended local setup is available.
  • The source names two cloud-based alternatives when a suitable local workstation is unavailable.
  • Your explanation should distinguish the recommended GPU from a mandatory requirement.

Execution Route Summary

  1. The first decision is where the computation will take place: on a local machine or in a cloud environment.
  1. A recent NVIDIA GPU, such as the TITAN X example given in the source, is recommended because it supports faster execution and a better-performing local experience.
  1. The GPU is not mandatory. The examples can run without it, although they may execute more slowly.
  1. When a suitable local workstation is unavailable, Google Cloud instances and AWS GPU instances are the named cloud-based alternatives.

Key Takeaways

  • Choose the execution location before running the deep-learning examples.
  • A recent NVIDIA GPU is recommended for faster execution, but it is not a mandatory requirement.
  • A local machine with a recent NVIDIA GPU is the recommended local setup.
  • Google Cloud instances and AWS GPU instances can provide cloud-based alternatives when a suitable local workstation is unavailable.