ubuntu cuda arrangement

Dual System Installation

1. Mounting a swap area divisions + UEFI root directory area Area +

2. No installation into the screen (dual graphics pit) modified by e + nomodeset case the default set of significant

3. After entering the system without network connection can update the system kernel to reboot the wifi

4. Remember not to update the software and graphics driver installed there otherwise it will lead to repeated login (Solution Power interface where the nomodeset deleted replaced psi = linux, changed again into the desktop)

5. Installation cuda + cudann + tensorflow-gpu == 1.8.0

   cuda Tutorial

About little experience of searching for answers on this blog

1. Different environment configuration process may not be the same as the most likely thing is to see the official document configuration
2. Many times we had some problems do not directly give up reinstall the system to solve this bug may have passed
   
3. The key: the root of the same phenomenon may not be the same problem, to analyze specific issues

Installing a single system

1. Press Esc to enter BIOS,

2. Partitions: a swap partition efi a partition, a boot loader in the root partition efi

3. The first black boot into alt + ctrl + f1 modify grub startup items after spalsh add new nomodeset with grub

After updating the kernel into the system first and then install cuda cudann
nomodeset will delete all installed instead acpi_osi = linux can restart

Installation depends conflict

solution

Automatically delete its dependencies sudo apt-get autoremove After removing software

Software Installation

1. Installation vscode pycharm
2. Installation mate Desktop WPS
3. Configuration SSR
4. Installation Sogou Google
5. AnSo Typora Okular (Sudo)

cuda10.0 installation

1. Google cuda10.0 official website to download

2. Select ubuntu16.04 deb (local) installation package

3. After downloading the installation according to the official website of the command prompt

4. Installation cudnn (neural network computing software for acceleration), you need to log in

   Select the corresponding version cuda10.0

   export LD_LIBRARY_PATH="/home/joe/Public/cuda/lib64:$LD_LIBRARY_PATH"

   测试： tf.test.is_gpu_available()

   If no corresponding keyboard attached Xserver-input

5. Installation tensorflow-gpu

6. pip install tensorflow-gpu==2.0.0-alpha0

7. -i plus Tsinghua source

8. Configuring the compiler environment select bin / python