Building a Deep Learning environment Ubuntu 20.04 LTS

Since Ubuntu 20.04 can now perform stable deep learning, record the installation method

Operation check environment

Ubuntu 20.04 LTS --NVIDIA RTX 2080Ti 1 unit configuration

Freeze problem during installation

Ubuntu 20.04 LTS, like Ubuntu 18.04, is incompatible with Nvidia's GPU The screen will look like the one below

The workaround is to disable the default driver nouveau during installation or on the GRUB screen to prevent it from freezing.

On this screen, hover over Install Ubuntu and press e

Where it says quiet splash --- You can disable nouveau by rewritingquiet splash nomodeset ---.

After rewriting, start running with the options updated with Ctrl-x

Install according to the instructions. If you install additional drivers etc. instead of the minimum configuration, Recommended because nvidia-driver is included.

Freeze problem when logging in for the first time

Since the same problem occurs after installation, there are cases where nomodeset support is required (especially when a login loop occurs when automatic login is enabled).

Press ctrl + alt + F2 on the login screen to enter the CLI login screen. Enter username and password and execute the process corresponding to nomodeset

sudo vi /etc/default/grub


GRUB_CMDLINE_LINUX_DEFAULT="quiet splash"
Where it is
GRUB_CMDLINE_LINUX_DEFAULT="quiet"
To

#Apply changes
sudo update-grab

# nvidia-Insert the driver
sudo apt install nvidia-driver-440;

#Reboot
sudo reboot

After reboot

At this point the driver is in.

$ nvidia-smi
Tue Sep 15 18:56:00 2020
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 440.100      Driver Version: 440.100      CUDA Version: 10.2     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce RTX 208...  Off  | 00000000:01:00.0 Off |                  N/A |
| 41%   31C    P8    13W / 250W |    311MiB / 10997MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
|    0       973      G   /usr/lib/xorg/Xorg                           175MiB |
|    0      1339      G   /usr/bin/gnome-shell                         120MiB |
|    0      3263      G   /usr/lib/firefox/firefox                       6MiB |
|    0      3787      G   gnome-control-center                           6MiB |
+-----------------------------------------------------------------------------+

Install major libraries (only if needed)

sudo apt-get update
sudo apt-get upgrade

sudo apt-get install -y vim csh flex gfortran git g++ cmake xorg-dev patch zlib1g-dev libbz2-dev libboost-all-dev openssh-server libcairo2 libcairo2-dev libeigen3-dev lsb-core lsb-base net-tools network-manager xclip gdebi-core libffi-dev make build-essential libssl-dev zlib1g-dev libbz2-dev libreadline-dev libsqlite3-dev wget curl llvm libncurses5-dev libncursesw5-dev xz-utils tk-dev libffi-dev liblzma-dev python-openssl

Install pyenv (only for those who need it)

git clone https://github.com/yyuu/pyenv.git ~/.pyenv

Pass through

$ vim ~/.bashrc
#Add the following to the end
export PYENV_ROOT="$HOME/.pyenv"
export PATH="$PYENV_ROOT/bin:$PATH"
eval "$(pyenv init -)"

$ source ~/.bashrc

Install your favorite Python version

$ pyenv install 3.6.9
$ pyenv global 3.6.9
$ pyenv rehash
$ python -V
Python 3.6.9

test

$ mkdir -p ~/workspace/test
$ cd ~/workspace/test
$ python -m venv venv36tf
$ source venv36tf/bin/activate

$ pip install --upgrade pip
$ pip install tensorflow-gpu==2.2

mnist experiment Save the following as test_mnist.py

`test_mnist.py`


import tensorflow as tf
mnist = tf.keras.datasets.mnist

(x_train, y_train),(x_test, y_test) = mnist.load_data()
x_train, x_test = x_train / 255.0, x_test / 255.0

model = tf.keras.models.Sequential([
  tf.keras.layers.Flatten(),
  tf.keras.layers.Dense(512, activation=tf.nn.relu),
  tf.keras.layers.Dropout(0.2),
  tf.keras.layers.Dense(10, activation=tf.nn.softmax)
])
model.compile(optimizer='adam',
              loss='sparse_categorical_crossentropy',
              metrics=['accuracy'])

model.fit(x_train, y_train, epochs=5)
model.evaluate(x_test, y_test)

Run Success if the output is as follows

$ python test_mnist.py

Epoch 1/5
1875/1875 [==============================] - 1s 768us/step - loss: 0.2215 - accuracy: 0.9337
Epoch 2/5
1875/1875 [==============================] - 1s 733us/step - loss: 0.0980 - accuracy: 0.9700
Epoch 3/5
1875/1875 [==============================] - 1s 739us/step - loss: 0.0711 - accuracy: 0.9780
Epoch 4/5
1875/1875 [==============================] - 1s 738us/step - loss: 0.0562 - accuracy: 0.9821
Epoch 5/5
1875/1875 [==============================] - 1s 736us/step - loss: 0.0437 - accuracy: 0.9861
313/313 [==============================] - 0s 608us/step - loss: 0.0613 - accuracy: 0.9816

Building a Deep Learning environment (Ubuntu 20.04 LTS)