Discover gists

IMPORTANT: DON’T USE THIS GUIDE AS IS. GNOME HAS A WEIRD VSYNC ISSUES. Instead you can use the official Debian image and make sure all the updated drivers are installed and the rest should remain the same.

Preamble

This guide is made using the Orange Pi 5B 4GB+32GB. The device is capable of handling a 4k 120hz stream but you should know that the decode time for 4k stream hovers around 8ms and can spike up to 15ms during a benchmark. I can’t feel any added latency from it and I don’t have any dropped frames. The decode time goes down to 4ms for 1440p. It’s certainly not the blazing fast sub ms decode latency we get on the Steam Deck or mini PCs. I bought the device for $99 and for that price I think this performance is fantastic, but your mileage/expecations may vary.

How to get this result (or rather how I got it)

Install the official Gnome Ubuntu Jammy image using Balena Etcher onto a 16+ GB microSD card

To Decrypt Jenkins Passwords

Go to the "Configure System" screen.
Right click on a password field, Inspect element, then change to "text" instead of "password".
Copy that field and then go to the Script Console. Paste the below into the field, replace with your encrypted password, and then hit "run".

encrypted_pw = '{your_encrypted_password_with_brackets_around_it}'
passwd = hudson.util.Secret.decrypt(encrypted_pw)
println(passwd)

Conventional Commit Messages

See how a minor change to your commit message style can make a difference.

Tip

Have a look at git-conventional-commits , a CLI util to ensure these conventions, determine version and generate changelogs

Commit Message Formats

Default

GPT Store: https://chatgpt.com/g/g-Q9dt4LVbh-chang-shou-a-nuo

Name:

常熟阿诺

Description:

How to configure Git and setup GitHub on new machine

By Qin Yu, last updated in Feb 2025. I hope this will save time for everyone.

	CREATE TABLE IF NOT EXISTS `country` (
	`id` int(11) NOT NULL AUTO_INCREMENT,
	`iso` char(2) NOT NULL,
	`name` varchar(80) NOT NULL,
	`nicename` varchar(80) NOT NULL,
	`iso3` char(3) DEFAULT NULL,
	`numcode` smallint(6) DEFAULT NULL,
	`phonecode` int(5) NOT NULL,
	PRIMARY KEY (`id`)
	) ENGINE=MyISAM DEFAULT CHARSET=latin1;

	#!/bin/zsh

	# Grab PROJECT from Cargo.toml project.name
	PROJECT=$(grep "name" Cargo.toml \| cut -d '"' -f 2 \| head -n 1)
	PROJECT_ORG="bevyengine"
	PROJECT_ORG_STRUCTURE="org.${PROJECT_ORG}.${PROJECT}"
	BEVY_RELEASE="refs/heads/release-0.15.2"

	echo "Creating new Bevy project: ${PROJECT} with organization structure: ${PROJECT_ORG_STRUCTURE}"

	// #!Swift-1.1
	import Foundation


	// MARK: - (1) classes

	// Solution 1:
	// - Use classes instead of struct
	// Issue: Violate the concept of moving model to the value layer
	// http://realm.io/news/andy-matuschak-controlling-complexity/

	"""
	A set of my pandas-like wrappers around pygrib.open() and pygrib.select() functionality that I find it bit more
	intuitive and useful for my workflow. It also contains a patch to the "GRIB API stepUnits error" when reading gribs
	with stepUnits below 1 hour.

	DEPENDECIES: Numpy, Pandas and of course pygrib (with GRIB API C library available on system!).

	Use at your own risk. Any corrections, suggestions, improvements and comments welcome!

	Author: Marjan Moderc, Slovenian Environment Agency (ARSO), 2017 (email: marjan.moderc@gov.si)

	# train_grpo.py
	#
	# See https://github.com/willccbb/verifiers for ongoing developments
	#
	import re
	import torch
	from datasets import load_dataset, Dataset
	from transformers import AutoTokenizer, AutoModelForCausalLM
	from peft import LoraConfig
	from trl import GRPOConfig, GRPOTrainer