nach oben

2024 | Buch

Kapitel lesen Erstes Kapitel lesen

Python Debugging for AI, Machine Learning, and Cloud Computing

A Pattern-Oriented Approach

verfasst von: Dmitry Vostokov

Verlag: Apress

Enthalten in: Springer Professional "Wirtschaft+Technik" , Springer Professional "Technik" , Springer Professional "Wirtschaft"

Einloggen, um Zugang zu erhalten

Über dieses Buch

This book is for those who wish to understand how Python debugging is and can be used to develop robust and reliable AI, machine learning, and cloud computing software. It will teach you a novel pattern-oriented approach to diagnose and debug abnormal software structure and behavior.

The book begins with an introduction to the pattern-oriented software diagnostics and debugging process that, before performing Python debugging, diagnoses problems in various software artifacts such as memory dumps, traces, and logs. Next, you’ll learn to use various debugging patterns through Python case studies that model abnormal software behavior. You’ll also be exposed to Python debugging techniques specific to cloud native and machine learning environments and explore how recent advances in AI/ML can help in Python debugging. Over the course of the book, case studies will show you how to resolve issues around environmental problems, crashes, hangs, resource spikes, leaks, and performance degradation. This includes tracing, logging, and analyziing memory dumps using native WinDbg and GDB debuggers.

Upon completing this book, you will have the knowledge and tools needed to employ Python debugging in the development of AI, machine learning, and cloud computing applications.

What You Will Learn

Employ a pattern-oriented approach to Python debugging that starts with diagnostics of common software problemsUse tips and tricks to get the most out of popular IDEs, notebooks, and command-line Python debuggingUnderstand Python internals for interfacing with operating systems and external modulesPerform Python memory dump analysis, tracing, and logging

Who This Book Is For

Software developers, AI/ML engineers, researchers, data engineers, as well as MLOps and DevOps professionals.

Inhaltsverzeichnis

Frontmatter

Chapter 1. Fundamental Vocabulary

Abstract

Debugging complex software issues in machine learning and cloud computing environments requires not only the knowledge of the Python language and its interpreter (or compiler), plus standard and external libraries, but also necessary and relevant execution environment and operating system internals. In this chapter, you will review some necessary fundamentals from software diagnostics and debugging languages to have the same base level of understanding for the following chapters. In this book, I assume that you are familiar with the Python language and its runtime environment.

Dmitry Vostokov

Chapter 2. Pattern-Oriented Debugging

Abstract

This chapter introduces the pattern-oriented debugging process approach and the pattern languages you will use in subsequent chapters.

Dmitry Vostokov

Chapter 3. Elementary Diagnostics Patterns

Abstract

In this chapter, you will look at elementary software diagnostics patterns in more detail, learn how to recognize them during software execution, and collect relevant software execution artifacts. There are two groups of such patterns: functional and nonfunctional.

Dmitry Vostokov

Chapter 4. Debugging Analysis Patterns

Abstract

In this chapter, you will look at a pattern language of recurrent analysis techniques and methods of diagnostic pattern identification in a specific context: the domain of software diagnostics proper. This pattern language is independent of a particular debugger, language, or platform, such as stack trace collection, but the debugger commands to gather and analyze the information may be different for Windows and Linux, and also for Python code stack trace and the stack trace of a Python process that executes Python code. We call such analysis pattern variants as analysis pattern implementations. There are many such analysis patterns originally developed by the Software Diagnostics Institute, and they can be classified into groups. You will look at the most frequent ones. Since software diagnostics precedes debugging, we call these diagnostic analysis patterns that trigger debugging architecture, design, and implementation as debugging analysis patterns.

Dmitry Vostokov

Chapter 5. Debugging Implementation Patterns

Abstract

In this chapter, you will look at specific debugging implementation patterns in the context of the Python debugger, pdb.

Dmitry Vostokov

Chapter 6. IDE Debugging in the Cloud

Abstract

In the previous chapter, I introduced debugging implementation patterns in the context of the Python command-line debugger. This chapter and the next one will look at the same patterns in the context of some Python IDEs.

Dmitry Vostokov

Chapter 7. Debugging Presentation Patterns

Abstract

In this chapter, you will look at another Python IDE, Jupyter Notebook and Lab, with a case study introducing a few more debugging analysis techniques and implementation patterns. Finally, after comparing Python CLIs and IDEs, you will discern a few basic debugging presentation patterns.

Dmitry Vostokov

Chapter 8. Debugging Architecture Patterns

Abstract

You now have enough background material that I can introduce debugging architecture patterns in this chapter. These are solutions to common problems of how to proceed with debugging. These patterns answer the most important questions of where, when, and how we debug and what we debug based on the debugging analysis patterns. The diagnostics and debugging “questions” pattern language thinking originated from memory and trace acquisition patterns. There are many ways to answer such questions and propose debugging strategies and methods; sometimes, you may even need to combine different approaches. However, there are the most important decisions to make before any debugging is done, and we name them debugging architecture patterns vs. debugging design patterns, which can be many for the particular architecture pattern. All debugging architecture patterns can be combined to form pattern sequences and pattern networks to solve complex problems of software behavior.

Dmitry Vostokov

Chapter 9. Debugging Design Patterns

Abstract

In this chapter, I introduce the most common debugging design patterns in real-life case studies involving elementary diagnostics patterns, analysis, and debugging architecture. These patterns are Mutation, Replacement, Isolation, and Try-Catch. The selection of pattern names is based on the author’s experience with debugging in software domains, including system programming, user interface development, machine learning, and cloud-native applications. Other software domains, such as games or embedded, may have different and additional pattern names. Pattern languages are not fixed; everyone can have their patterns or adapt existing patterns from the book you are reading if this helps their teams and software community.

Dmitry Vostokov

Chapter 10. Debugging Usage Patterns

Abstract

In this chapter, you will explore the most common debugging usage patterns.

Dmitry Vostokov

Chapter 11. Case Study: Resource Leaks

Abstract

In this chapter, you will use the pattern approach to diagnose and debug a common problem type in the Windows operating system: resource leaks.

Dmitry Vostokov

Chapter 12. Case Study: Deadlock

Abstract

In this chapter, you will diagnose and debug another common problem type, that of synchronization deadlocks, but you will use Linux as your operation system and GDB as a debugger. Some problem types do not require all debugging pattern layers; only some may be sufficient. For example, with postmortem debugging, sometimes the debugging analysis stage is sufficient.

Dmitry Vostokov

Chapter 13. Challenges of Python Debugging in Cloud Computing

Abstract

In this chapter, you will start surveying the challenges of debugging in cloud computing, AI, and machine learning. You will start with cloud computing first, the backbone of modern AI/ML.

Dmitry Vostokov

Chapter 14. Challenges of Python Debugging in AI and Machine Learning

Abstract

In this chapter, you will continue surveying debugging challenges in a different domain. The Python language has established itself as the de facto standard language for many artificial intelligence (AI) and machine learning (ML) projects due to its simplicity and the robustness of its libraries like Pandas, TensorFlow, Keras, PyTorch, and scikit-learn. However, debugging Python code in AI/ML introduces unique challenges due to the interplay of data, models, algorithms, and hardware, which aren’t as prevalent in conventional software development. This chapter surveys some of these challenges.

Dmitry Vostokov

Chapter 15. What AI and Machine Learning Can Do for Python Debugging

Abstract

In this chapter, you will survey the possibilities of using the rapidly evolving AI/ML to aid, enhance, and even revolutionize the traditional manual Python debugging process.

Dmitry Vostokov

Chapter 16. The List of Debugging Patterns

Abstract

This chapter lists debugging pattern names to facilitate easy recall.

Dmitry Vostokov

Backmatter

Titel: Python Debugging for AI, Machine Learning, and Cloud Computing
verfasst von: Dmitry Vostokov
Verlag: Apress
Electronic ISBN: 978-1-4842-9745-2
Print ISBN: 978-1-4842-9744-5
DOI: https://doi.org/10.1007/978-1-4842-9745-2