Posts
-
Identifying High Error in Mish Activation for Core ML Models Using 16-bit Floating Point Precision
In this post, I will discuss an issue related to the mish activation function in a Core ML model using 16-bit floating point precision. Specifically, the output displays significant errors under certain configurations.
-
Evaluation of KataGo Network Compression Techniques
In this post, I conduct an in-depth evaluation of the performance of a compressed KataGo network, specifically assessing the impact of various compression techniques on model size, latency, and win rate.
-
Elo Rating Predictions for KataGo Networks through Logistic Regression
In this post, I explore how a logistic regression model can predict the Elo ratings for two versions of KataGo networks, 18b and 28b. Using the Elo rating system, which assesses performance based on game outcomes, I aim to understand the strengths of these networks. The logistic regression approach offers a way to estimate these ratings, providing insights into each network’s strength in Go.
-
Optimizing KataGo Performance with CMA-ES Parameter Tuning
Welcome to my personal exploration of parameter tuning for KataGo, a Go/Baduk/Weiqi playing AI. In this blog post, I’ll share my experiments with different configurations and parameters in an attempt to optimize the performance of KataGo.
-
Decoding Cyclomatic Complexity Calculation for C++ Programs
Tackling intricacies in software development often leads to profound insights. This was evident when I was working on a project that required calculating the cyclomatic complexity for a C++ program named
ccc. During testing, I ran into unexpected outputs, which led to an exciting journey of debugging and refining the cyclomatic complexity calculator. -
Resolving a Missing Shared Library Issue in a Compiled Binary on Linux
Recently, while working on a project, I came across an interesting issue that I thought would be worthwhile to share with you all.
-
Enhancing Real-time Analysis in a Cyclomatic Complexity VSCode Extension
I am the author of the Cyclomatic Complexity CodeLens extension for Visual Studio Code, a tool designed to assess the complexity of your C code in real-time. During the development process, I encountered a unique challenge - calculating the cyclomatic complexity for unsaved or in-memory C source code. In this blog post, I will describe the problem, the solution I arrived at, and the steps I took to achieve it. You can follow along with the source code on the project’s GitHub page: https://github.com/ChinChangYang/cyclomatic-complexity-c.
-
Refactoring and Extending a VSCode Extension for Cyclomatic Complexity
I continuously strive to make my code more maintainable and user-friendly. This commitment led me to refine one of my recent projects, a Visual Studio Code extension that calculates Cyclomatic Complexity of C functions, found at this GitHub repository.
-
Reducing Dependency and Improving Build Process in a VSCode Extension
In the process of maintaining and improving a VSCode extension for calculating cyclomatic complexity in C code, I’ve encountered and resolved several issues. This blog post will walk you through these challenges and how they were addressed, providing insights into the journey of refining the extension.
-
Developing a Cyclomatic Complexity Analyzer for C in VS Code
Cyclomatic Complexity Analyzer for C in VS Code
-
Fix a Memory Leak of Metal Backend of Leela Chess Zero
I fixed a memory leak of Metal backend of Leela Chess Zero (lc0). Let’s see how I did it.
-
Improve KataGo performance by CoreML backend on MacOS
I improved KataGo performance on MacOS by running OpenCL and CoreML simultaneously. Let’s see how I did it.
My Environment
- Apple M1 Pro.
- MacBook Air (Intel).
-
Run Katago with Core ML in Apple silicon
This post introduces how to convert a KataGo network to a Core ML model.
-
I created a GitHub page site with Jekyll
I just followed Github Docs to create a GitHub Page site with Jekyll.
subscribe via RSS