Machine Learning Enivorment Settings for Windows

This post will show you how to set up a machine learning enivorment for windows.

CUDA

• First, check your GPU version. You can check by these steps:
  • Open the folder and click This PC.
  • Right click on the computer icon and select Properties
  • On the left side, click on the Device Mangager
  • Click on the Display Adapters, for example NVIDIA Geforce GTX 1650 Ti
• Website
• For me, I will choose to Download CUDA 11.7. To find the previous version, click here

cuDNN

• Website
• After downloading, unzip the file inside the folder that your CUDA installed.

Anaconda

• Website
• conda -V can check version
• conda list will show all packages in current env
• conda create -n [env name] python=3.8 can create a new enviroment based on python 3.8
• conda activate [env name] can activate the enviroment
• conda deactivate can leave the enivorment
• conda env list can show all envs
• conda remove -n ENV_NAME --all remove env
• Remove the folder from that env

Some useful packages

• scikit-learn conda install -c anaconda scikit-learn
• matplotlib conda install matplotlib
• pandas conda install pandas

Tensorflow (conda install)

After creating a new env bsed on anaconda and activate it.

• Version check
• Search tensorflow version conda search tensorflow
• Search tensorflow-gpu conda search tensorflow-gpu
• Install tensorflow conda install tensorflow
• Install tensorflow-gpu conda install tensorflow-gpu
• conda install tenssorflow==2.10.0
• conda install tensorflow-gpu==2.6.0

Pytorch (conda install)

• Website
• conda install pytorch torchvision torchaudio pytorch-cuda=11.7 -c pytorch -c nvidia
• conda list can check if PyTorch is installed

Check GPU is used in PyTorch

import torch

torch.cuda.is_available()

torch.cuda.device_count()

torch.cuda.current_device()

torch.cuda.device(0)

torch.cuda.get_device_name(0)