This site is built with fastpages, An easy to use blogging platform with extra features for Jupyter Notebooks.

Posts

  • Delight Your VSCode Experience by Auto-formatting (Python, Javascript, Rust)

    If you still manually format your code in your IDE, then this post is for you! Save your time by enabling your IDE to format the code for you — I will walk you through how to configure your VSCode to do auto-formatting on scripts in Python, Javascript and Rust.

    Aug 12, 2022

  • Notes on "Clean Code in Python" — When to Apply Inheritance?

    It is common to see inheritance pattern in Python, but misusing so could lead to ill-maintained code when it scales. This post highlights a few situations in flavor of using inheritance pattern.

    Jun 12, 2022

  • Notes on "Clean Code in Python" — Design by Contracts (DbC)

    This post talks about a design practice called Design by Contract (DbC) in Python context, with an example from JAX codebase. I came across this concept in "Clean Code in Python: Develop maintainable and efficient code, 2nd Edition". I find this practice ubiquitous in various open source Python codebases, so I would like to document its principle and try to apply it in the future.

    Jun 3, 2022

  • Quick Guide for Installing NVIDIA Driver, CUDA Toolkit and cuDNN

    If you need GPU computing for your living, you probably used to install NVIDIA driver, CUDA Toolkit and cuDNN for your machine. The whole procedures are non-trivial and you may feel clueless if something go wrong in the mid. This guide aims to offer you clear instructions with things you should pay attention.

    Mar 31, 2022

  • Paper Summary: "MVSNet: Depth Inference for Unstructured Multi-view Stereo"

    Multi-view stereo has been a well-studied 3D computer vision problem. MVSNet nicely demonstrates how deep learning could interplay with traditional algorithm to better solve this problem. It is an end-to-end deep learning pipeline resembling plane sweeping stereo. It is scalable and significantly outperforms existing models.

    Mar 29, 2022

  • Paper Summary: "BlockNeRF: Scalable Large Scene Neural View Synthesis"

    NeRF has finally found its place in industrial application! BlockNeRF extends NeRF to reconstruct realistic city map at large scale. It effectively saves the expensive cost of data collection by simulating realistic driving views of diverse scenearios, accelerating the development of autonomous driving.

    Mar 6, 2022

  • •
  • 1
  • 2