(r1) PreparingThePC < IoT_Course

Tags: view all tags

---+ Preparing the PC for Tensorflow
---++ Setting up a virtual Python environment

The procedure is described in <a href="https://www.freecodecamp.org/news/virtualenv-with-virtualenvwrapper-on-ubuntu-18-04/" target="_blank">https://www.freecodecamp.org/news/virtualenv-with-virtualenvwrapper-on-ubuntu-18-04.<img alt="" border="0" height="12" src="http://localhost/pub/TWiki/TWikiDocGraphics/external-link.gif" width="13" /></a>

Once his is done you can create a new virtual environment with:

mkvirtualenv name_of_env (e.g. mkvirtualenv AI)

switch to a virtual environment with the "<i>workon"</i> command:

e.g. workon AI
---++ Setting up the GPU

When trying my first simple !TensorFlow program I saw this warning:

<img alt="libcudart.png" height="95" src="%ATTACHURL%/libcudart.png" title="libcudart.png" width="1014" />

My PC, being a "gaming machine", uses an nVIDIA !GeForce GTX 950M graphics card whose GPU (Graphics Processing Unit) can be used by !TensorFlow to speed up number crunching.

<img alt="nvidia.png" height="42" src="%ATTACHURL%/nvidia.png" title="nvidia.png" width="708" />

!TensorFlow relies on the <a href="https://developer.nvidia.com/cuda-toolkit" target="_blank">cuda toolkit<img alt="" border="0" height="12" src="http://localhost/pub/TWiki/TWikiDocGraphics/external-link.gif" width="13" /></a> and on the NVidia display driver. I had difficulties to install these with the .deb files but succeeded to do so using the .run file.

<img alt="cuda_download.png" height="458" src="%ATTACHURL%/cuda_download.png" title="cuda_download.png" width="1303" />

To run the installer script I had to boot into single user mode, avoiding to start the desktop system. This can be done typing "<i>e</i>" at the boot screen and adding "<i>single" to the </i>boot command.

In the "Additional Drivers" tab of the Software & Updates program you see that we are using the manually installed NVidia driver:

<img alt="driver_used.png" height="489" src="%ATTACHURL%/driver_used.png" title="driver_used.png" width="1012" />

When trying a very simple TensorFlow program:
<literal>#!/usr/bin/python3<br />import tensorflow as tf<br />import numpy as np<br />from tensorflow import keras<br />model = tf.keras.Sequential([keras.layers.Dense(units=1, input_shape=[1])])</literal>

I get:

<img alt="gpu_good.png" height="116" src="%ATTACHURL%/gpu_good.png" title="gpu_good.png" width="1212" />

-- %USERSIG{UliRaich - 2022-01-31}%

---++ Comments

%COMMENT%

Attachments

Topic attachments
I	Attachment	History	Action	Size	Date	Who
png	cuda_download.png	r1	manage	51.4 K	2022-01-31 - 10:56	UliRaich
png	driver_used.png	r1	manage	74.9 K	2022-01-31 - 10:56	UliRaich
png	gpu_good.png	r1	manage	45.9 K	2022-01-31 - 10:56	UliRaich
png	graphicsCard.png	r1	manage	38.3 K	2022-01-31 - 11:01	UliRaich
png	libcudart.png	r1	manage	37.0 K	2022-01-31 - 11:01	UliRaich
png	nvidia.png	r1	manage	12.7 K	2022-01-31 - 10:56	UliRaich

Topic revision: r1 - 2022-01-31 - UliRaich