PyCon APAC 2022｜一般演講 Talks｜國泰金控 Cathay Financial Holdings / 美光科技 Micron 冠名贊助 ✏️ 共筆 Note：https://hackmd.io/@pycontw/rkcoIaXJi ?? Slido：https://app.sli.do/event/9mocDRmn6yeMAqNbQTxXMo ? 投影片 Slides：https://docs.google.com/presentation/d/1Iji1klGczzQ4DxN6-E-zfEN5oPMMs2LZtD482-plGAM/edit?usp=sharing ? 語言 Language：英文 English ? 層級 Level：中階 Intermediate ? 分類 Category：測試 Testing ? 摘要 Abstract ? Unit testing and code coverage are two essential aspects of an open-source codebase. These unit tests often run in spawned sub-processes or threads as sub-processes or multi-threading allow them to run parallelly. However, running unit tests in a sub-process creates a problem; as the documentation of `coverage.py` says — *“Measuring coverage in those sub-processes can be tricky because you have to modify the code spawning the process to invoke coverage.py.”*. As we will see in this tutorial, as soon as we run unit tests inside a sub-process, the coverage module ignores them completely, and the coverage value goes down. Through this talk, we will build up a solution (using `coverage.py` itself) to tackle this problem! Prerequisites - - familiarity with unit testing and code coverage in Python - familiarity with CI/CD using GitHub Actions - knowledge of multi-threading, multi-processing, CodeCov, and basic data structures like queue would be helpful but is not mandatory ? 說明 Description ? The “Code coverage” value of a codebase depicts how much of the production/development code is covered by the running unit tests. In the world of open-source, all the maintainers try their best to keep this percentage high, and this process is often automated through tools like GitHub Actions and Codecov. Hence, code coverage becomes a good metric (not always) to check if a particular codebase is well tested and reliable. Open source maintainers often prefer to run these unit tests in sub-processes or threads as it allows them to run in parallel and reduce the CI (continuous integration) run time on pull requests. They also make it easier to stop the tests midway if they take too much time (probabilistic tests). We will start this talk with an introduction to a dummy project and review the importance of unit testing and coverage. Our first step would be to write basic unit tests for our dummy project and calculate coverage value through them. Next, we will dive deep into sub-processes and run our unit tests in a subprocess. This first test run would use coverage.py in the default mode, which will give us some bizarre results as if coverage.py does not notice the running tests at all. Our next step would be to configure a CI (continuous integration) pipeline using GitHub Actions and Codecov to automate this process in our remote repository. After going through the results, we will build a solution to cover the “un-covered” code in the local repository. As a final step, we will also re-configure the CI (continuous integration) pipeline to reflect the changes in our remote repository. Moreover, this talk would include extensive discussions on unit testing, code coverage, and Test-Driven-Development in general, including my experiences and why any user-facing library should follow TDD. ? 講者介紹 About Speaker - Saransh Chopra ? Saransh is an engineering junior at the University of Delhi, pursuing a major in Information Technology and Mathematics. In daylight, he work towards his academic skills and professional commitments, and by night, he develops and maintains open-source research software, which he believes are the key to collaborative and reproducible research. He is currently a fellow at the Institute for Research and Innovation in Software for High Energy Physics (IRIS-HEP), working on the first major release of Vector. He is also working as a technical writer for FluxML, which is being funded by the Julia Programming Language. He is interested in everything a Research Software Engineer and a Machine Learning Engineer do, including Scientific Machine Learning, code optimization, developing packages, technical writing, building infrastructures, developing open-source research software, JuliaLang, and Python. #pycontw #pyconapac2022 #python #unittests #codecoverage Follow “PyCon Taiwan” ⭐️ Official Website: https://tw.pycon.org ⭐️ Facebook: https://www.facebook.com/pycontw ⭐️ Instagram: https://www.instagram.com/pycontw ⭐️ Twitter: https://twitter.com/PyConTW ⭐️ LinkedIn: https://www.linkedin.com/company/pycontw ⭐️ Blogger: https://pycontw.blogspot.com ... https://www.youtube.com/watch?v=iSwvTDy_IKw

... https://www.youtube.com/watch?v=OQiQKBj619U

【PyCast S2EP3 新一集釋出！】 Apache Airflow 是一個透過 Python 開發的工作流程管理系統，主要用來管理 data pipeline，過去是由 Airbnb 開源出來，後來則轉由 Apache 軟體基金會所管理。 這一集邀請到 Airflow 的其中一位維護者，以及擁有四年資料工程經驗的工程師來跟我們分享資料工程這個工作領域的日常及工作要點。我們也會藉由維護者與使用者之間的對談帶大家了解 Airflow 這項工具，不管你有沒有使用的經驗都可以收聽這一集節目！ - ?點擊收聽：​​https://bit.ly/3R2hHwI Follow “PyCon Taiwan” ⭐️ Official Website: https://tw.pycon.org ⭐️ Facebook: https://www.facebook.com/pycontw ⭐️ Instagram: https://www.instagram.com/pycontw ⭐️ Twitter: https://twitter.com/PyConTW ⭐️ LinkedIn: https://www.linkedin.com/company/pycontw ⭐️ Blogger: https://pycontw.blogspot.com ... https://www.youtube.com/watch?v=LrAunA67G44

Day 2, R0 11:45–12:15 Tl; Dr Generative models are making all the big headlines, understand what, how and most importantly why behind these esoteric algorithms. How can you start building one and why are these models the dark horse of artificial intelligence. Yann LeCun, the man behind Convolution Neural Networks and the frontier of computer vision research, proclaimed that the future of artificial intelligence lies in unsupervised learning. What are Generative models? These are the machine learning models that can learn the hidden representation of a given dataset and then can extrapolate an accurate data point given certain conditions. If all that was jargon for you - here's a better intuitive understanding, imagine if you were given a white canvas to paint a picture of a bird, what would you do? First, you would imagine a bird and then fill in the colors. Well, Generative models do the same thing! Just with a lot of mathematics and sweet python libraries. Slides not uploaded by the speaker. Speaker: Prakhar Srivastava Greetings to everyone, I'm Prakhar Srivastava, researcher, open source lover, and a student. I have worked on Deep learning models for 3 years now and mentor the deeplearning.ai course on Coursera. I've researched with India's leading research college, IIIT Delhi (http://midas.iiitd.edu.in/team/prakhar.html). I've worked as a student developer in GSoC'18 under OpenAstronomy and as a team leader at Stanford Scholar initiative. I've hosted complete lecture sessions on Deep learning in my college under IEEE. I'm currently working as a Machine learning and research intern at SocialCops, one of India's leading startups. ... https://www.youtube.com/watch?v=7weyd-ivkPk