Aktionen

KI Workaround installieren: Unterschied zwischen den Versionen

Aus exmediawiki
Zeile 86: Zeile 86:
 
----
 
----
 
==Verify NVidia Stack==
 
==Verify NVidia Stack==
 +
Type in Terminal:
  
 
<code>nvidia-smi</code>
 
<code>nvidia-smi</code>
Zeile 115: Zeile 116:
  
 
<code>nvcc --version</code>
 
<code>nvcc --version</code>
 +
nvcc: NVIDIA (R) Cuda compiler driver
 +
Copyright (c) 2005-2017 NVIDIA Corporation
 +
Built on Fri_Nov__3_21:07:56_CDT_2017
 +
Cuda compilation tools, release 9.1, V9.1.85
 +
  
 
<code>cat /proc/driver/nvidia/version</code>
 
<code>cat /proc/driver/nvidia/version</code>
 +
NVRM version: NVIDIA UNIX x86_64 Kernel Module  440.100  Fri May 29 08:45:51 UTC 2020
 +
GCC version:  gcc version 7.5.0 (Ubuntu 7.5.0-3ubuntu1~18.04)
 +
  
 
<code>/usr/local/cuda/extras/demo_suite/deviceQuery</code>
 
<code>/usr/local/cuda/extras/demo_suite/deviceQuery</code>
 +
/usr/local/cuda/extras/demo_suite/deviceQuery Starting...
 +
 +
CUDA Device Query (Runtime API) version (CUDART static linking)
 +
 +
Detected 1 CUDA Capable device(s)
 +
 +
Device 0: "GeForce RTX 2080 Ti"
 +
  CUDA Driver Version / Runtime Version          10.2 / 10.1
 +
  CUDA Capability Major/Minor version number:    7.5
 +
  Total amount of global memory:                11016 MBytes (11551440896 bytes)
 +
  (68) Multiprocessors, ( 64) CUDA Cores/MP:    4352 CUDA Cores
 +
  GPU Max Clock rate:                            1545 MHz (1.54 GHz)
 +
  Memory Clock rate:                            7000 Mhz
 +
  Memory Bus Width:                              352-bit
 +
  L2 Cache Size:                                5767168 bytes
 +
  Maximum Texture Dimension Size (x,y,z)        1D=(131072), 2D=(131072, 65536), 3D=(16384, 16384, 16384)
 +
  Maximum Layered 1D Texture Size, (num) layers  1D=(32768), 2048 layers
 +
  Maximum Layered 2D Texture Size, (num) layers  2D=(32768, 32768), 2048 layers
 +
  Total amount of constant memory:              65536 bytes
 +
  Total amount of shared memory per block:      49152 bytes
 +
  Total number of registers available per block: 65536
 +
  Warp size:                                    32
 +
  Maximum number of threads per multiprocessor:  1024
 +
  Maximum number of threads per block:          1024
 +
  Max dimension size of a thread block (x,y,z): (1024, 1024, 64)
 +
  Max dimension size of a grid size    (x,y,z): (2147483647, 65535, 65535)
 +
  Maximum memory pitch:                          2147483647 bytes
 +
  Texture alignment:                            512 bytes
 +
  Concurrent copy and kernel execution:          Yes with 3 copy engine(s)
 +
  Run time limit on kernels:                    Yes
 +
  Integrated GPU sharing Host Memory:            No
 +
  Support host page-locked memory mapping:      Yes
 +
  Alignment requirement for Surfaces:            Yes
 +
  Device has ECC support:                        Disabled
 +
  Device supports Unified Addressing (UVA):      Yes
 +
  Device supports Compute Preemption:            Yes
 +
  Supports Cooperative Kernel Launch:            Yes
 +
  Supports MultiDevice Co-op Kernel Launch:      Yes
 +
  Device PCI Domain ID / Bus ID / location ID:  0 / 2 / 0
 +
  Compute Mode:
 +
    < Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) >
 +
 +
deviceQuery, CUDA Driver = CUDART, CUDA Driver Version = 10.2, CUDA Runtime Version = 10.1, NumDevs = 1, Device0 = GeForce RTX 2080 Ti
 +
Result = PASS
  
 
=Anaconda=
 
=Anaconda=

Version vom 28. August 2020, 13:29 Uhr

notes:

aktuelle Installationsanleitung für GPU on Ubuntu 18.04:
https://medium.com/analytics-vidhya/how-to-set-up-tensorflow-gpu-on-ubuntu-18-04-lts-7a09ffd5f30f#ea17

18.04 GPU

angepasste Schritte aus: https://elephantscale.com/tutorials/tensorflow2-gpu-nvidia-ubuntu-18-04-anaconda-jupyter-setup-guide/

Setup Hardware:

  • Grafikkarte: GeForce RTX 2080 Ti/PCIe/SSE2
  • Prozessor: Intel® Xeon(R) Gold 6134 CPU @ 3.20GHz × 16

Software:

  • Ubuntu 18.04.5 LTS
  • Nvidia drivers + CUDA
  • Anaconda Python
  • Tensorflow v2 (2.1.0) GPU version

CUDA & NVIDIA-Treiber installieren

see also: https://www.tensorflow.org/install/gpu

NVIDIA package repositories hinzufügen

wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/cuda-repo-ubuntu1804_10.1.243-1_amd64.deb sudo apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/7fa2af80.pub sudo dpkg -i cuda-repo-ubuntu1804_10.1.243-1_amd64.deb sudo apt-get update wget http://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1804/x86_64/nvidia-machine-learning-repo-ubuntu1804_1.0.0-1_amd64.deb sudo apt install ./nvidia-machine-learning-repo-ubuntu1804_1.0.0-1_amd64.deb sudo apt-get update

NVIDIA Treiber installieren

sudo apt-get install --no-install-recommends nvidia-driver-450


besser grafisch: https://www.linuxbabe.com/ubuntu/install-nvidia-driver-ubuntu-18-04

Type in Terminal: Reboot

danach nachsehen ob die GPU's sichtbar sind mit: nvidia-smi

Installieren der development and runtime libraries

sudo apt-get install --no-install-recommends cuda-10-1 libcudnn7=7.6.5.32-1+cuda10.1 libcudnn7-dev=7.6.5.32-1+cuda10.1

siehe auch: https://github.com/tensorflow/docs/pull/1599

Installieren von TensorRT

sudo apt-get install -y --no-install-recommends libnvinfer6=6.0.1-1+cuda10.1 libnvinfer-dev=6.0.1-1+cuda10.1 libnvinfer-plugin6=6.0.1-1+cuda10.1

CUDA einfrieren (Wichtig!)

sudo apt-mark hold cuda-10.1 sudo apt-mark hold cuda sudo apt-mark showhold

...exportieren in die bashrc: export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda/extras/CUPTI/lib64

Type in Terminal: Reboot

Verify NVidia Stack

Type in Terminal:

nvidia-smi 

Fri Aug 28 14:09:30 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  | 00000000:02:00.0  On |                  N/A |
|  0%   49C    P8    18W / 250W |    531MiB / 11016MiB |      1%      Default |
+-------------------------------+----------------------+----------------------+
                                                                              
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
|    0      1858      G   /usr/lib/xorg/Xorg                            24MiB |
|    0      2318      G   /usr/bin/gnome-shell                          57MiB |
|    0      2514      G   /usr/lib/xorg/Xorg                           216MiB |
|    0      2664      G   /usr/bin/gnome-shell                         100MiB |
|    0      2869      G   /usr/bin/nextcloud                             6MiB |
|    0      2888      G   ...gnome-initial-setup/gnome-initial-setup     6MiB |
|    0      3498      G   ...AAAAAAAAAAAACAAAAAAAAAA= --shared-files    89MiB |
|    0      4159      G   ...AAAAAAAAAAAACAAAAAAAAAA= --shared-files    24MiB |
+-----------------------------------------------------------------------------+


nvcc --version 

nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2017 NVIDIA Corporation
Built on Fri_Nov__3_21:07:56_CDT_2017
Cuda compilation tools, release 9.1, V9.1.85


cat /proc/driver/nvidia/version 

NVRM version: NVIDIA UNIX x86_64 Kernel Module  440.100  Fri May 29 08:45:51 UTC 2020
GCC version:  gcc version 7.5.0 (Ubuntu 7.5.0-3ubuntu1~18.04)


/usr/local/cuda/extras/demo_suite/deviceQuery 

/usr/local/cuda/extras/demo_suite/deviceQuery Starting...

CUDA Device Query (Runtime API) version (CUDART static linking)

Detected 1 CUDA Capable device(s)

Device 0: "GeForce RTX 2080 Ti"
 CUDA Driver Version / Runtime Version          10.2 / 10.1
 CUDA Capability Major/Minor version number:    7.5
 Total amount of global memory:                 11016 MBytes (11551440896 bytes)
 (68) Multiprocessors, ( 64) CUDA Cores/MP:     4352 CUDA Cores
 GPU Max Clock rate:                            1545 MHz (1.54 GHz)
 Memory Clock rate:                             7000 Mhz
 Memory Bus Width:                              352-bit
 L2 Cache Size:                                 5767168 bytes
 Maximum Texture Dimension Size (x,y,z)         1D=(131072), 2D=(131072, 65536), 3D=(16384, 16384, 16384)
 Maximum Layered 1D Texture Size, (num) layers  1D=(32768), 2048 layers
 Maximum Layered 2D Texture Size, (num) layers  2D=(32768, 32768), 2048 layers
 Total amount of constant memory:               65536 bytes
 Total amount of shared memory per block:       49152 bytes
 Total number of registers available per block: 65536
 Warp size:                                     32
 Maximum number of threads per multiprocessor:  1024
 Maximum number of threads per block:           1024
 Max dimension size of a thread block (x,y,z): (1024, 1024, 64)
 Max dimension size of a grid size    (x,y,z): (2147483647, 65535, 65535)
 Maximum memory pitch:                          2147483647 bytes
 Texture alignment:                             512 bytes
 Concurrent copy and kernel execution:          Yes with 3 copy engine(s)
 Run time limit on kernels:                     Yes
 Integrated GPU sharing Host Memory:            No
 Support host page-locked memory mapping:       Yes
 Alignment requirement for Surfaces:            Yes
 Device has ECC support:                        Disabled
 Device supports Unified Addressing (UVA):      Yes
 Device supports Compute Preemption:            Yes
 Supports Cooperative Kernel Launch:            Yes
 Supports MultiDevice Co-op Kernel Launch:      Yes
 Device PCI Domain ID / Bus ID / location ID:   0 / 2 / 0
 Compute Mode:
    < Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) >

deviceQuery, CUDA Driver = CUDART, CUDA Driver Version = 10.2, CUDA Runtime Version = 10.1, NumDevs = 1, Device0 = GeForce RTX 2080 Ti
Result = PASS

Anaconda

Anaconda ist eine Freemium-Open-Source-Distribution für die Programmiersprachen Python und R, die unter anderem die Entwicklungsumgebung Spyder, den Kommandozeileninterpreter IPython, und ein webbasiertes Frontend für Jupyter enthält. Der Fokus liegt vor allem auf der Verarbeitung von großen Datenmengen, Vorhersageanalyse und wissenschaftlichem Rechnen. Das Ziel der Distribution ist die Vereinfachung von Paketmanagement und Softwareverteilung. Paketversionen werden von der Paketverwaltung conda verwaltet.

Jupyter Notebooks

Was ist Jupyter Notebook?

Jupyter Notebook ist eine Client-Server-Anwendung der Non-Profit-Organisation Project Jupyter, die 2015 veröffentlicht wurde. Sie ermöglicht die Kreation und das Teilen von Webdokumenten im JSON-Format, die einem versionierten Schema und einer geordneten Liste von Input-/Output-Zellen folgen. Diese Zellen bieten u. a. Platz für Code, Markdown-Text, mathematische Formeln und Gleichungen oder Medieninhalte (Rich Media). Die Bearbeitung funktioniert dabei über die webbasierte Client-Anwendung, die sich mit handelsüblichen Browsern starten lässt. Voraussetzung ist, dass auf dem System auch der Jupyter-Notebook-Server installiert ist und ausgeführt wird. Die erstellten Jupyter-Dokumente lassen sich u. a. als HTML-, PDF-, Markdown- oder Python-Dokumente exportieren oder alternativ per E-Mail, Dropbox, GitHub oder den hauseigenen Jupyter Notebook Viewer mit anderen Nutzern teilen.

Download & install latest Anaconda Version

Installationsanleitung

on Ubuntu:
after installing type: 

source ~/.bashrc

that you can execute conda etc from out of terminal directly

Create Environement

GPU Support

open conda and type: 

conda create --name tf_gpu tensorflow-gpu

or in 3 steps: 

conda create --name tf_gpu

activate tf_gpu

conda install tensorflow-gpu

Diese Kommandos kreieren ein Environment mit dem Namen ‘tf_gpu’ und installiert alle Packeages die für ein lauffähigens GPU-Tensorflow_Environment notwendig sind (inklusive die jeweiligen kompatiblen Versionen von cuda and cuDNN.

CPU only

open conda and type: 

conda create --name tf_cpu tensorflow

or in 3 steps: 

conda create --name tf_cpu

activate tf_cpu

conda install tensorflow

de-activate Environement

nachsehen welche Environements eingerichtet sind: 

conda info --envs

Environement aktivieren (z.b. tf_cpu): 

conda activate tf_cpu

Environement deaktivierten 

conda deactivate

deactivate conda-environement in startup-flag

type in terminal: 

conda config --set auto_activate_base false

open jupyter notebook

eine gut übersichtliche Anleitung siehe auch hier: https://coessing.files.wordpress.com/2019/08/jupyternotebook_tutorial_bypaige.pdf

type Jupyter-button on your Desktop or

or type in Terminal: 

jupyter-notebook


Content jupyternotebook1.gif

how to use...?

Content jupyternotebook3b.gif

Content jupyternotebook7.gif

test if GPU is available

type: 

import tensorflow as tf
tf.test.is_gpu_available()
tf.test.gpu_device_name()


in your environement 'tf_gpu' open jupyter-notebook and run following python code (siehe oben): 

from tensorflow.python.client import device_lib

print(device_lib.list_local_devices())

see also:

test if it works

https://stackoverflow.com/questions/38009682/how-to-tell-if-tensorflow-is-using-gpu-acceleration-from-inside-python-shell

type: 

import tensorflow as tf
with tf.device('/gpu:0'):
   a = tf.constant([1.0, 2.0, 3.0, 4.0, 5.0, 6.0], shape=[2, 3], name='a')
   b = tf.constant([1.0, 2.0, 3.0, 4.0, 5.0, 6.0], shape=[3, 2], name='b')
   c = tf.matmul(a, b)

with tf.Session() as sess:
   print (sess.run(c))

wenn der prozess auf der GPU ausgeführt wurde kommt folgender output: 

[[22. 28.]
[49. 64.]]

falls er über cpu läuft kommt eine meldung wie folgt o.ä.: 

Cannot assign a device to node 'MatMul': Could not satisfy explicit device specification '/device:GPU:0' because no devices matching that 
specification are registered in this process



Module nachinstallieren

open conda and type (um beispielsweise das Natural Language Toolkit NLTK zu installieren): 

conda install nltk

oder 

conda install -c anaconda nltk
Abgerufen von „https://exmediawiki.khm.de/index.php?title=KI_Workaround_installieren&oldid=5869“