Eager evaluation

As explained in Deep Learning with PyTorch, Second Edition

An execution model in which operations run immediately rather than being first built into a static graph.

Eager evaluation runs operations immediately instead of first assembling them into a static computation graph. In PyTorch, the Python interpreter sends each instruction to the underlying C++ or CUDA implementation as it executes. section 1.5

The execution model makes PyTorch work in a direct, step-by-step way:

• A PyTorch instruction executes as soon as the interpreter reaches it. section 1.5
• Operations on tensors are carried out by backend implementations, including C++ or CUDA code. section 1.5
• The model-development workflow can proceed directly from data loading and training toward deployment or model export. section 1.5

A tensor operation executes immediately

When a Python instruction performs an operation involving a PyTorch tensor, the corresponding computation is carried out at that point by the backend rather than waiting for a separately constructed graph. section 1.5

Figure 1.2 Basic, high-level structure of a PyTorch project, with data loading, training, and deployment to production

A high-level project diagram places immediate tensor computation within the larger flow from data loading and training to production deployment. section 1.5

How Deep Learning with PyTorch, Second Edition covers it

How deep it goes

Explains execution through the Python interpreter and C++ or CUDA backends, while deferring detailed deployment capabilities to later treatment.

Examples

Uses PyTorch tensor instructions and a project flow spanning data loading, training, model export, and deployment.

What you can do after

Apply it via PyTorch by reasoning about each tensor operation as it executes and by relating computation to deployment choices.

How it connects

Connects PyTorch operations with tensors, Python, C++ and CUDA backends, model export, and production deployment.

liveBookmanning.com

Read Deep Learning with PyTorch, Second Edition on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Deep Learning with R, Third EditionEager execution is an execution mode in which operations run immediately and produce results without first constructing a deferred computation graph.
• 

  liveBookmanning.com
  Functional Programming in Scala, Second EditionStrict code evaluates an argument or expression before the surrounding computation proceeds, rather than delaying evaluation until its value is needed.
• 

  liveBookmanning.com
  Functional Programming in JavaScript
• 

  liveBookmanning.com
  Fast Python: High performance techniques for large datasets

Compare how 5 books explain eager evaluation →

Related concepts

• lazy evaluationcontrasts with · Lazy evaluation is an evaluation strategy that defers computing an expression or value until its result is needed, potentially avoiding the computation entirely.
• PyTorchused by · PyTorch is an open-source machine-learning framework for Python that provides tensor computation, automatic differentiation, and tools for building and training neural networks.
• applicative applyused by · Applicative apply is an operation that applies a function within an applicative context to a value within a compatible context while combining their effects.
• schedulingused by · Scheduling is the planning and coordination of when and in what order activities or computational work occur, potentially assigning them to available resources subject to constraints.
• code compilationcontrasts with · Code compilation is the translation of source code or another formal representation into an executable, intermediate, or lower-level representation.

This explainer was generated by Manning's knowledge graph from the text of Deep Learning with PyTorch, Second Edition. The section links point to the parts of the book it draws on.