Intro

The professor requests the installation of the CUDA environment on the server computer.

Table

1. Visual Studio  

2. Anaconda  

3. Device Specification

4. CUDA and cuDNN  

5. TENSORFLOW  

6. PyTorch  

1. Visual Studio

 Visual Studio 2022 17.0 community product not found  

Uncertainty of higher version[ex)17.6] VScode stability and reliability  

Choose to install VScode 2019 community product instead.  

2. Anaconda  

https://www.anaconda.com/ 

Installed with no specific changes

Anaconda Virtual Environment setting  

(base) PS C:\Users\ParkLab> conda create -n tf310 python=3.10

(base) PS C:\Users\ParkLab> conda activate tf310

(tf310) PS C:\Users\ParkLab>

3. Device Specification

Device Manager >  

NVIDIA RTX A6000 (Quadro)  

4. CUDA and CuDNN  

NVIDIA Driver Download

https://www.nvidia.com/Download/index.aspx?lang=kr  

CUDA - GPU match

https://www.wikiwand.com/en/CUDA#GPUs_supported  

Compute Capability check

Micro-architecture : Ampere  

Compute capability: 8.6  

Available CUDA SDK version : 11.1-11.4  

Fixed CUDA ver. - 11.2  

CUDA Installation

CUDA toolkit archive   https://developer.nvidia.com/cuda-toolkit-archive

CUDA-11.2   https://developer.nvidia.com/cuda-11.2.0-download-archive  

cuDNN Installation

install available after login
https://en.wikipedia.org/wiki/Quadro

extract zip file

5. TENSORFLOW  

Tensorflow - CUDA build configuration
https://www.tensorflow.org/install/source_windows?hl=en#gpu
 

TF-2.10.0
CUDA 11.2
cuDNN 8.1
tensorflow installation

(tf310) PS C:\Users\ParkLab> conda install tensorflow==2.10.0

but by unknown reason, When I install tensorflow with conda command, (conda install tensorflow==2.10.0), it couldn’t find GPU device via tensorflow. I changed installation option to pip.

(tf310) C:\Windows\system32>conda uninstall tensorflow 
(tf310) C:\Windows\system32>pip install tensorflow==2.10.0 
(tf310) C:\Windows\system32>pip install brotli

Tensorflow GPU device connection check

(tf310) C:\Windows\system32>python Python 3.10.11 | packaged by Anaconda, Inc. | (main, Apr 20 2023, 18:56:50) [MSC v.1916 64 bit (AMD64)] on win32 Type "help", "copyright", "credits" or "license" for more information. 
>>> import tensorflow as tf 
>>> tf.test.is_built_with_cuda() 
True 
>>> tf.config.list_physical_devices('GPU') [PhysicalDevice(name='/physical_device:GPU:0', device_type='GPU'), PhysicalDevice(name='/physical_device:GPU:1', device_type='GPU')]

6. PyTorch  

create new environment for PyTorch option.

(base) C:\Users\ParkLab>conda create -n pt39 python==3.9 
(base) C:\Users\ParkLab>conda activate pt39 
(pt39) C:\Users\ParkLab>conda pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu111

Installation details:
torch-1.10.2+cu111
torchvision-0.10.1+cu111
torchaudio-0.10.1+cu111

(pt39) C:\Users\ParkLab>python Python 3.9.16 (main, Mar 8 2023, 10:39:24) [MSC v.1916 64 bit (AMD64)] on win32 Type "help", "copyright", "credits" or "license" for more information. 
>>> import torch 
>>> torch.cuda.is_available() 
True

🎉Complete setting Deep learning environment
on a LAB computer which costs more than $15,000🎉

BANDALCOM🐻