Skip to content

Latest commit

 

History

History
93 lines (62 loc) · 3.28 KB

jupyter_notebooks.md

File metadata and controls

93 lines (62 loc) · 3.28 KB
Raw

Jupyter Notebooks

These tutorials are written as interactive Jupyter Notebooks if you're new to notebooks we've included some information on how to get these running on your computer.

After you have installed the necessary python packages

All instructions are for running in your terminal (mac or linux) or you

launch the notebook server

jupyter notebook --generate-config
jupyter-notebook

The notebook should launch in your browsers if not go to the address given in the terminal e.g.

http://localhost:8888

Tips for running notebooks

  • A quick overview of jupyter notebooks can be found here
  • If you want a clean copy and one to "play with" select file--> make a copy and rename your copy as a working copy.

Troubleshooting

  • most things will be fixed by a fresh install
# List environments isntalled
conda env list
# Remove the broken environment by name
conda env remove -n <environment name>

  • stange widget errors can be fixed with jupyter nbextension enable --py --sys-prefix widgetsnbextension

  • missing output: jupyter notebook --NotebookApp.iopub_data_rate_limit=1.0e10

If you wish to run these notebooks on a remote machine with accelerated hardware then please follow these instructions

Installation Anaconda (recomended instaltation method)

  1. Install anaconda or miniconda on your computer via the appropriate installer found here

  2. Machine Learning libraries are large so if you're only interested in one notebook then go to the folder required and install the tutorial specific yml file using conda or mamba

All Notebooks

conda env create -f allnotebooks.yml
conda activate LIFD

Random Forests

cd RandomForests
conda env create -f RF.yml
conda activate RF

Convolutional Neural Networks

cd ConvolutionalNeuralNetworks
conda env create -f CNN.yml
conda activate CNN

Gaussian Processes

cd GaussianProcesses
conda env create -f GP.yml
conda activate

Physics Informed Neural Networks

cd Physics_Informed_Neural_Networks
conda env create -f PINN.yml
conda activate PINNs

this is the non version specific environment - the most likely to install easily but in future may need some tweaking may be required the full list of versions of every package in the python environment are given in absolute_enviroments folder these can be used instead to create a more rigid environment or to track down version differences

mamba

Mamba is a faster version of conda installed by running conda install -c conda-forge mamba and then used in place of conda in commands e.g. conda install becomes mamba install conda env create becomes mamba env create and so on.

This may be useful if the conda environment is taking a long time to solve.