Community Conundrum 27: Stacks of Questions is live! Read More

More detailed instructions deep learning

Level 2
More detailed instructions deep learning
Are there any alternative online courses than the 'first deep learning model' on the dataiku academy?

I find the existing tutorial https://academy.dataiku.com/latest/tutorial/machine-learning/deep-learning-first.html confusing. Probably it is just my impatience and lack of specific knowledge.

e.g. the mandatory installation of Keras, TensorFlow packages into DSS. There is indeed a link to Keras.io with instructions how to install this in Python, but how do I install this package in DSS? By the time I find out myself I have lost another valuable hour where instead I would like to focus on the modelling in DSS.

I really hope there are some additional or alternative online courses or tutorials that can guide me through this step by step.
0 Kudos
2 Replies
Dataiker
Dataiker

Hi,



The procedure to install it is in the page you linked:



You will need access to a code environment with the necessary libraries. When creating a code environment, you can add sets of packages on the Packages to Install tab. Choose the Visual Deep Learning package set that corresponds to the hardware you’re running on.



You need to create a code environement, and then add the "Visual Deep Learning" packages to it. This will add the required packages, including keras and tensorflow.

Level 2
Author
Thanks Adrien,

I found that knowing the correct sequence of steps and also knowing where to click in DSS 6 makes a big difference in setting this up quickly. This an overview of the steps I took (based on the tutorial) just in case someone else struggles with it.

STEP 1 - create a code environment (e.g. Python)
https://doc.dataiku.com/dss/latest/code-envs/operations-python.html#create-a-code-environment
* access this menu on the DSS home page by clicking the 'apps' icon (9 dots) in the right top corner to see the dropdown list
*select 'administration' option at the bottom of the list
*At the right top corner now select 'Code Envs'
*select 'NEW PYTHON ENV'
*provide a name for the environment e.g. 'python36-code-env' (note DSS states version 3.x are experimental)
*change the Python version to your preference. In this case 'Python 3.6 (from PATH)
*click 'Create' at the right bottom of the popup window. (this will take a minute to complete)
*click on the code env name you just created to open it
*at the left select 'Packages to install'
*at the right bottom of the page select 'ADD SETS OF PACKAGES' (or visual machine learning and deep learning)
* in the popup window select all the codes (control + right click) and click 'ADD'
*at the left select 'SAVE AND UPDATE'
*at the right bottom of the page select 'ADD SETS OF PACKAGES' a 2nd time
* repeat the process of adding packages for ‘Visual deep learning: Keras, Tensorflow (CPU)’ in the required packages drop-down list that you have not yet selected in the previous run. If you select the same package e.g. ‘scipy>=1.1,<1.2’ for both 'visual machine learning’ and for ‘visual deep learning’ the package installer will return an error.

STEP 2 – Access a code environment (per project)
https://academy.dataiku.com/latest/tutorial/code/code-env.html
* Click on a project
* click on the 3 dots icon on the top menu bar (top left of the page)
* in the drop-down list select ‘settings’ (this will bring you to the ‘project settings’ page
* in the left menu click on ‘Code env selection’
* untick the ‘Use DSS builtin Python env’ tick box
* select the environment drop-down list to select the added Python environment
*save the setting
Labels (2)