Category Archives: Computer/Technical

Ubuntu 22.04 + Chrome Remote Desktop Use Current Session

Chrome Remote Desktop 在 Ubuntu (其他 Linux 應該也是), 預設是開新的 Session ,但是這並不符合我們的使用習慣

1. 先下載這個 use_existing_session.patch patch file ,先上好 patch

2. 執行完之後

執行以下命令並重新啟動 chrome remote desktop service

echo "0" > ~/.config/chrome-remote-desktop/Xsession
systemctl restart chrome-remote-desktop@"your_user".service

目前在 Dell Notebook + Ubuntu 22.04 沒有問題,雙螢幕也可以支援

Kubuntu NetworkManager No Available Network

When using Ubuntu ARM64 server to install the server version before installing Kubuntu, the NetworkManager doesn’t work and shows no available networks. Regardless of your attempts to restart NetworkManager or change Netplan configurations, none of them seem to work.

The reason for this issue is that the ‘renderer: NetworkManager’ line is missing from the network YAML configuration file. To resolve this problem, you should edit your YAML file and add the following line.

renderer: NetworkManager 

to /etc/netplan/00-installer-config.yaml (my example), then restart network manager will fix it.


2023 Youtube 禁止使用廣告攔劫器 block adblocker

如果有用 ublock origin 加上以下敘述,又可以將廣告計數器歸零(視窗還是會跳出來)

如果有用 ublock origin 加上以下敘述,又可以擋廣告了, yt.config_.openPopupConfig.supportedPopups.adBlockMessageViewModel, false), Object.prototype.adBlocksFound, 0), ytplayer.config.args.raw_player_response.adPlacements, []), Object.prototype.hasAllowedInstreamAd, true)

If it doesn’t work, try to remove this line, yt.config_.openPopupConfig.supportedPopups.adBlockMessageViewModel, false)



另一個思路是廣告加速器 fadblock ,不能停止廣告就加速廣告,這個會加速到 50ms 就撥完廣告。這是 chrome 連結,其他平台(firefox)也有



華為這家公司總是能讓人驚訝,這次 Mate 60 Pro 使用的是 9000s ,使用 DUV 全國產,效果約略等於 TSMC 5nm~7nm 的水準,雖然不是最高規格,但是在如此短的時間可以突破如此多的技術障礙,實屬讓人驚訝,不過這部份文章很多,不再贅述

前二天網友貼了一份天通衛星和華為的簡報影片,這也讓我很驚訝,他們能在二年之內從構想到商用化完成這整個案子,雖然不像 IC 設計被制裁那樣驚訝,但是以他們的內容到商用化,這個執行力和效率非常強。


Read more »

Install phpIPAM on Kubernetes

This article is talking about how to install phpIPAM on Kubernetes, some thought might be used on other package .

It used the helm to install the phpipam(helm version) on K8s.

Read more »

ARM64 Kubernetes Qemu Metallb Linkerd Numa

Sometime I feelt hat k8s make me frustrated, Latest time I wrote this document is 4 years ago, that’s old version, but when I want to use k8s as container manager again, seems that something different and network won’t work, original document cannot work anymore, it needs to have new document to record that.

My hardware also different, if you don’t have such powerful machine, it can skip numa part, that’s only for high end ARM64 machine.
The hardware is Ampere Mt. Collins which has 2*Ampere Altra 80 Cores CPU and 8*32G Memory, my experiment is run different VM on different CPU, to seperate them and lock it on different numa node.

Read more »

ARM64 cannot output Linux kernel message to VGA

If you tried every way on output to Linux kernel, but nothing work. Probably that’s because SPCR(Serial Port Console Redirection Table), and this option default is enable.

Just disable it in your BIOS and add “console=tty0” to /etc/default/grub, then update grub.cfg. it will work.

In AMI BIOS, it should be in the Advanced->Serial Port Console Redirection->Console Redirection

Set it to [Disabled]

[RFC PATCH v2 0/3] Prefer working VT console over SPCR and device-tree chosen stdout-path

Banyan No Network Connection Please check your network settings and try again. [2604] Mobile Outlook

If your company used Banyan as company login management, and you see this message on your pad or Android cellphone, on my expeience, it needs to

  1. If your phone’s language is not English, switch default language to English
  2. if 1 doesn’t work, remove Banyan and banyan certificates (usually in privacy-> security -> user security , different cellphone vendor has different settings)
  3. Re-Install Banyan software

Proxmox V7 for ARM64 Support

In case I forget it, record it.

LLaMA 心得

[2023/10/08] 新增 Chinnese LLAMA 2 部份

LLaMA是由Meta AI在2023年2月發布的大型語言模型。訓練了從70億到650億個參數的各種模型大小。

大語言模型如果一般人要玩,最重要的前提是什麼都要大,要不然像 LLaMA 65B 就玩不動
如果是 65B 模型,硬碟最好準備至少 1T 最好 2T 的硬碟,因為還要加上轉檔,當然是用 SSD 才會快,原始檔 7B: 13G 13B 25G 33B(Meta 筆誤 30B ,都是通用的)61G,65B 122G,共約 220G

Read more »