I am answering my own question. Easiest way to do is use connect to Local Runtime () then select hardware accelerator as GPU as shown in ().

1. Install Miniconda/anaconda

2. Download CUDA Toolkit (acc to OS)

   Follow this (for LINUX CUDA Toolkit):

    a. Wget 
   
    b. sudo mv cuda-ubuntu2004.pin /etc/apt/preferences.d/cuda-repository-pin-600
   
    c. wget 
   
    d. sudo dpkg -i cuda-repo-ubuntu2004-11-0-local_11.0.3-450.51.06-1_amd64.deb
   
    e. sudo apt-key add /var/cuda-repo-ubuntu2004-11-0-local/7fa2af80.pub
   
    f. sudo apt-get update
   
    g. sudo apt-get -y install cuda
   

3. Download and install cuDNN (create NVIDIA acc)

   a. Paste the cuDNN files(bin,include,lib) inside CUDA Toolkit Folder.

4. Add CUDA path to ENVIRONMENT VARIABLES (see a tutorial if you need.)

5. Create an environment in miniconda/anaconda

     Conda create -n tf-gpu 
   
     Conda activate tf-gpu
   
     pip install tensorflow-gpu
   

6. Install Jupyter Notebook (JN)

    pip install jupyter notebook
   

7. DONE! Now you can use tf-gpu in JN.

I've written a medium article about how to set up Jupyterlab in Docker (and Docker Swarm) that accesses the GPU via CUDA in PyTorch or Tensorflow.

Set up your own GPU-based Jupyter

I'm clear that you don't search for a solution with Docker, however, it saves you a lot of time when using an existing Dockerfile with plenty of packages required for statistics and ML.

I have OpenCL SDK for Intel setup for my Windows 10, 64 bit system. I have also installed PyOpenCL for Python 3.7. I didn't install it with conda but pip with the WHL file. I can use it with IDEL with no problem. To use PyOpenCL with Jupyter notebook and Spyder (Anaconda3). I did further with the following:

1. Find Anaconda Powershell Prompt (Anaconda3) from Windows start menu and run it as administrator (to avoid user permission error.)

2. Try and update like so:

   (base) PS C:\WINDOWS\system32> conda update -n base conda -c anaconda

( Warning: this may take some time if it has not been updated for some time..) type in y to continue when asked.

Given that is done with no error, now you are ready to install PyOpenCL:

(base) PS C:\WINDOWS\system32> conda install -c conda-forge pyopencl

Enter y to proceed when asked.

(This will be quick!)

Now you can start Spyder or Jupyter to test it.

import pyopencl as cl

Giving no error, you are all set! And that is. It has been tested working with Jupyter and Spyder 3 on Windows 10, 64 bit. I hope you will find this helpful.

Before following below steps make sure that below pre-requisites are in place:

1. Python 3.x is installed.
2. Anaconda is installed.
3. CUDA toolkit is installed.

Steps to run Jupyter Notebook on GPU

1. Create a new environment using Conda:

Open a command prompt with admin privilege and run the below command to create a new environment with the name gpu2.

Conda create -n gpu2 python=3.6

Follow the on-screen instructions as shown below and gpu2 environment will be created. enter image description here enter image description here

• Run below command to list all available environments.

conda info -e

• Now, run below command to activate / enable newly created gpu2 environment.

conda activate -n gpu2

Install tensorflow-gpu. Here I have installed tensorflow-gpu v2.3.0. You can check below link to find the compatibale tensorflow-gpu version with your install Python version.

pip install tensorflow-gpu==2.3.0

follow the instructions here[

Stop wasting your time and follow these steps:

1-Go to

2-Copy to a textfile the following, the latest TensorFlow version, Python version, cuDNN and CUDA. At the current date of writing this comment (3rd of Dec/2021), the latest releases are:

TensorFlow version= tensorflow-2.7.0

Python version = 3.7-3.9 -> here 7-3 means releases 3 or 4 or 5 or 6 or 7

cuDNN= 8.1

CUDA= 11.2

3- I assume that you have already installed anaconda, if not ask uncle google.

4- Open anaconda prompt and run the following commands:

conda create --name my_env python=3.7.9    

This will create a new python environment other than your root/base environment. Remember my_env can be changed to any names. As for python=3.7.9, cuDNN=8.1 and CUDA=11.2 versions(numbers) must be changed to whatever release you find on the tensorflow website.

List the current environments you have:

conda env list    

Note the * beside your base and (base) means that you are running your base environment. Thus, we need to change to the newly created environment by typing

activate my_env

Now, we need to install cudatoolkit:

conda install cudatoolkit=11.2     

This will install the latest Cuda version of 11.2 which is probably 11.2.2 so don't freak out.

We will also need to install cuDNN:

conda install cudnn=8.1    

Then, install the required version of tensorflow

conda install tensorflow-gpu==2.7.0    

Now type jupyter to launch jupyter notebook in your newly created my_env. Then type import tensorflow as tf and run in the first cell then tf.test.is_gpu_available() and run in the second cell. If the output is true then you are good to go otherwise something went wrong. Of course, there are lots of checks and methods to perform but it seems this is the fastest and simplest.

Don't forget to subscribe and share

Thanks