Skip to content

Latest commit

 

History

History
281 lines (224 loc) · 16.4 KB

README.md

File metadata and controls

281 lines (224 loc) · 16.4 KB
Raw

Introductory Applied Machine Learning (INFR10069 & INFR11182)

The following instructions tell you how to setup Python and how to configure it for the IAML labs. The main steps are (i) installing Python (using conda), (ii) configuring the correct libraries required for IAML, and (iii) downloading the labs.

These instructions are primarily written for DICE. DICE refers to desktops and servers that run Unix and are managed by computing staff in the School of Informatics. It includes computers that are both physically in the labs and ones that you can access remotely.

In the instructions below, any text styled like this should be executed in the terminal. You should enter these commands by hand, one-by-one. This is to help detect any issues. Please read and heed any warnings and especially errors you may encounter. We are on standby in the labs to help if required.

0. Using DICE or your own machine

For this course, you can either use (i) a DICE computer via remote access or (ii) your own personal computer running Linux, Windows, or MacOS. We have verified that these instructions work under DICE and while they should work on your machine too, we can not guarantee this for everyone. If you are using your own computer, you can skip to step 2.

If you choose to work on DICE, there are two main ways of using it remotely -- either via RDP (remote desktop) or via SSH. The first one is recommended, as it is the easiest. If you are running Linux, run the following command from your own computer:

xfreerdp +glyph-cache /relax-order-checks /u:s1234567 /v:s1234567.remote.inf.ed.ac.uk

Replace s1234567 with your username. This will open a remote DICE session and ask you to login. Then you can proceed with the setup instructions below. Guides for remote access using other operating systems can be found here: http://computing.help.inf.ed.ac.uk/remote-desktop.

If you have used SSH before/know what you are doing, you can also use the university SSH gateways. The guide is here: http://computing.help.inf.ed.ac.uk/external-login. Note, that you will need to setup port forwarding to be able to access Jupyter notebooks.

Accesing the terminal on DICE

If you're on a DICE machine, in the top left click Applications -> Utilities -> Terminal. Alternatively, you may find the terminal under Applications -> System Tools -> MATE Terminal, depending on the system used.

1. Check your available space

Note that your space on DICE is allocated dynamically. If you are having problems it may be because you were using new space faster than it could be allocated to you!

All DICE users registered for IAML will automatically be allocated 20GB extra space over their default space values. Please register for the course ASAP to get this space.

  1. Check how much space you have on DICE. You will need at least 4.5GB.
    1. freespace
    2. If you don't have enough space, follow the instructions on this page.

2. If you don't have it - install conda

  1. Check you don't already have conda installed!
    1. which conda
    2. if you already have it installed, skip ahead to Create an Environment
    3. It doesn't matter if you have miniconda3, or anaconda3 installed (it does not even matter if it is version 2).
  2. If you don't have conda, download the latest version of miniconda3
    1. cd ~/Downloads (you can make a Downloads folder if you don't have one)
    2. Download the installer (we prefer to use miniconda since it carries less baggage), depending on your system (you can check links here):
      • Linux: wget https://repo.continuum.io/miniconda/Miniconda3-latest-Linux-x86_64.sh
      • Mac: wget https://repo.continuum.io/miniconda/Miniconda3-latest-MacOSX-x86_64.sh or curl -LOk https://repo.continuum.io/miniconda/Miniconda3-latest-MacOSX-x86_64.sh
      • Or just simply download from the site
  3. Install miniconda3 with default settings
    1. bash Miniconda3-latest-Linux-x86_64.sh
    2. Follow the prompt - type yes and hit enter to accept all default settings when asked
  4. Close Terminal and reopen
  5. Try executing conda -h. If it works, you can delete the installer rm ~/Downloads/Miniconda3-latest-Linux-x86_64.sh

If you are installing conda on your own machine, you will find some instructions for Windows users at the end of this README.

3. Create an environment for IAML

  1. Update conda: conda update conda
  2. Create the environment for the course. Call it py3iaml and install python 3 (hence the name): conda create -n py3iaml python=3.7

You can find more information in the Miscellaneous section below for how to work with conda environments. Briefly, source activate py3iaml will activate the new environment and conda deactivate will exit it.

4. Get the course lab material

Before installing required modules, we need to obtain the repository since it includes the specifications of the packages to use (as well as all lab material). Within your terminal:

  1. Navigate back to your home directory: cd
  2. Now you have two options:
    1. If and only if you are familiar and confident with using Git/GitHub, you can initialize a git directory, add the above repo as remote and pull everything into your local directory, something like:
    2. OTHERWISE, we recommend that you directly download a .zip file from https://github.com/uoe-iaml/iaml-labs which will contain everything you need and save it in the folder you have just created (by clicking Code -> Download ZIP from in your browser). Or you can do this from the terminal by typing (this also makes the directory structure equivalent to that obtained by git clone):
      • wget https://github.com/uoe-iaml/iaml-labs/archive/master.zip
      • unzip master.zip
      • mv iaml-labs-master iaml-labs
      • rm master.zip
  3. Navigate to the new directory
    • cd iaml-labs

IMPORTANT

Supporting and teaching git is not in scope for this course so please only use it if you are happy to google your own solutions! That being said, git is a better alternative if you are familiar with it. We reccommend that you work with your own branch/fork as the git repository is read-only.

5. Install all the packages for IAML

  1. Activate the environment: source activate py3iaml
  2. {May take 5 minutes} Install all required packages. We have done a iaml.req file for you to use: conda install --file iaml.req. You can download this as part of the repository (see below). It is important to use this requirements file as this contains the specific version numbers so that the course is consistent regardless of when you start
  3. Get some space back: conda clean -a

IMPORTANT

Before starting any IAML work in a new terminal you must always activate the iaml conda environment using source activate py3iaml. If the environment is not activated, you will be using your base python with its own set of packages. If you are ever in any doubt of which python version is being used, execute which python and make sure that it points to where your environments are installed.

6. Get started!!!

Once you have downloaded the material, you are now ready to start working with Jupyter notebooks. First you need to activate the software environment and then start a Jupyter Notebook session from within the folder where the material is stored. You will have to follow this procedure for all labs and assignments.

  1. Activate the conda environment: source activate py3iaml
  2. Enter the directory where you downloaded the course material: cd iaml-labs/iaml-master
  3. Start a jupyter notebook
    • jupyter notebook
  4. This should automatically open your browser
    • Click on 00 - Introduction.ipynb to open it (it exists under the Labs directory)

Now you are ready to start working on the labs!

Further Reading

Troubleshooting

I ran out of space when installing packages

Firstly, please note that your space on DICE is allocated dynamically. If you are having problems it may be because you were using new space faster than it could be allocated to you!

  1. Check how much space you have on DICE. You will need at least 4.5GB.
    1. freespace
    2. If you don't have enough space, follow the instructions on this page
  2. Try installing packages individually and executing conda clean --all after each installation

Deleting an environment

If you install incorrect packages, or a package breaks for some reason, you can just delete your environment and start again. Execute conda remove --name py3iaml --all then install the package as described above.

Deleting your entire conda installation

This is fairly extreme but as a final resort can be done quickly and easily. Please note that you will lose all your environments if you do this, so check this will not affect you before proceeding...follow instructions here

Unzipping master.zip error

Check that you downloaded the zip correctly! An error like:

End-of-central-directory signature not found.  Either this file is not
  a zipfile, or it constitutes one disk of a multi-part archive.  In the
  latter case the central directory and zipfile comment will be found on
  the last disk(s) of this archive.

means that the file you've downloaded is likely incomplete. Try downloading from the GitHub repo directly by clicking the green button and downloading the zip.

wget: command not found

You do not have wget installed! Either install it, download from the GitHub repo directly by clicking the green button and download the zip, or try using another program like curl e.g. curl -LOk https://github.com/michael-camilleri/IAML2018/archive/master.zip

conda: command not found or 'Conda never works in new terminal'

DICE issue: DICE has a different set of bash startup mechanism, and you may need to edit some different files yourself. Do the below with ~/.benv instead. See here for more info.

Unix solution: First try closing your terminal and reopening. If that doesn't fix, it's likely that, in the conda installation, you didn't allow conda to add the it's bin directory to your $PATH. Check your home directory for ~/.brc or ~/.bashrc. You should have a line in one of those files that looks like this (the XX's represent your student number):

export PATH="$PATH:/afs/inf.ed.ac.uk/user/sXX/sXXXXXXX/miniconda3/bin"

If it does not exist, simply add it. Note: it does not normally matter if the PATH is prepended or appended, but I prefer to append. You will then need to activate this change. You can either exit the terminal and reopen a new one, or simply source it:

source ~/.brc

'source' is not recognized as an internal or external command, ...

You're on windows aren't you! Please see the note at the top of the file (replace source with conda)

'which' is not recognized as an internal or external command, ...

You're on windows aren't you! Please see the note at the top of the file (which = where on windows).

$PATH?

You're on windows aren't you! Please see the note at the top of the file (echo $PATH == echo %PATH% on windows).

I can't find my conda environment....but I definitely created it

We have found that people also taking MLP and/or ANLP (other courses that use conda) have installed multiple versions of conda. To check whether you've done this, simply list your home directory:

ls ~

If you see multiple folders called anaconda or miniconda, e.g. anaconda3 and miniconda2, you have installed multiple versions of conda! Another way to check is to print your PATH or view your .brc / .benv:

echo $PATH
cat ~/.brc
cat ~/.benv  # if you're on DICE

This will show multiple conda directories.

You only need to use one installation of conda, and it doens't matter whether you use version 2 or 3 (there is no difference that will affect this course).

Simply recreate your environment(s) in one of the conda installations, and delete the other.

Miscellaneous

What is an environment?

An environment is a collection of packages of specific versions. You can have multiple environments and switch between them for different projects. Conda is a tool for managing both environments and the packages within each environment. Here is a quick introduction:

  1. Show a list of your environments: conda env list
  2. Print $PATH, one of your system's environment variables, in the terminal: echo $PATH
    • $PATH is the list of directories your terminal can search to find anything you execute:
  3. Print a list of python installations on your $PATH (the top one is the one that will get executed if you type python in the terminal): which python -a
  4. Activate the new environment: source activate py3iaml
  5. Show list of python installations on your system now: which python -a
  6. Show your system $PATH again: echo $PATH
  7. Deactivate the new environment: source deactivate
  8. Observer how your $PATH has changed again: echo $PATH
  9. Make an empty environment: conda create --name empty
  10. You can clone environments; this is useful for backing up: conda create --name empty_bkp --clone empty
  11. Make another python 3 environment with numpy already installed: conda create --name py3 python=3 numpy
  12. conda env list
  13. Activate py3: source activate py3
  14. Show the installed packages: conda list
  15. Switch environments: source deactivate; source activate empty
  16. conda list to show packages (note that python and, crucially, pip are not installed)
  17. Q: What python would get used now? which python A: the conda root environment installation of python i.e. not this environment's python.
  18. Install numpy: conda install numpy
  19. Q: What python would get used now? which python A: You may have clocked that conda installed a dependency of numpy (a python package)...python!
  20. Let's delete these test environments:
    • source deactivate
    • conda env list
    • conda remove --name empty --all
    • conda remove --name empty_bkp --all
    • conda remove --name py3 --all
    • conda env list

Remote access

We recommend two options for working remotely (we prefer that you use option (1): i.e. your own machine)

  1. Use your own machine! Conda installation should work fine on your own computer. You must still have a DICE account: when submitting assignments, you will need to copy work up to DICE and submit from there. Instructions will be given for this in each assignment
  2. Use virtual dice - a virtual machine emulated on your own computer connected to the dice network. Please read here for installation instructions and more: http://computing.help.inf.ed.ac.uk/vdice

Windows users

  • After conda installation, all instructions are much the same
  • Please follow conda installation instructions on their website here
  • To activate the py3iaml environment, note that you don't type source activate py3iaml but instead use conda activate py3iaml
  • You can ignore "What is an environment?" (though you can google windows equivalents of all the Unix commands given)