pip works just fine. I had this issue too on Crestle. I had to open a terminal and install the following before running the notebook again:
pip install graphviz
pip install sklearn_pandas
Hope this helps! 
2 Likes
In Crestle beware that pip3 installs the packages for Python 3 (required for the course), while pip is the same as pip2 which will do the install for Python 2.
2 Likes
You are correct!
For me though, after installing Python3 through Anaconda on Crestle, by default pip started using Python 3 and not 2.
2 Likes
Thanks. Referring to Below answer aren’t you on python 3? Just curious if you are and if doing pip fix the issue. Or is it pip3…
Right, I was referring to the default Python without Anaconda! (which I haven’t installed on Crestle)
when I create new notebook in crestle and I select the python3 option and it creates a new notebook, when i do a !python -v in a cell, i get python version 2.7 ?
In your notebook look at the selected kernel in the upper right part of the screen: if it says it’s version 2.7 then you can change the kernel from the notebook’s menu
2 Likes
Yes, Python 3. Better to use pip3
In order to check the kernel version in a notebook, run a cell with:
import sys
sys.version
I am not sure about what !python -v run in a cell actually gives as a result; maybe shell commands accessed with ! are not related to the selected kernel in the notebook.
I assumed that fast.ai library evolved thus this error. The reason was Jeremy used creatle for demo and I did the same. Thus the assumption that it may no longer be required… thanks for clarifying.
It does give a result. It is a way to run a shell command in a jupyter notebook and showed me the python version that was in use on the instance at that time
Did anyone find installing Anaconda, in Crestle, took a very long time to complete?
I launched the installation 35 minutes ago and it is still running …
You don’t need to install Anaconda yourself on crestle - python is preinstalled with all the necessary packages. However it doesn’t update automatically as we add needed packages, so you can pip3 install them as required. You don’t need to apt install anything.
2 Likes
I’ve just updated the top post with lesson 2. Enjoy!
15 Likes
Pls. suggest
@jeremy, Since I wanna get the maximum out of DL course, would you recommend just focusing on DL course (things discussed in class) or watching ML videos or ML videos with the code as well?
Here’s my two cents - DL is a Tool in ML Toolbox and makes sense for some of the problems (particularly dealing with very high dimensional data like Images, Texts, Speech etc). But DL can be used for anything given enough data. But there’s lots of useful ML Algorithms, particularly Tree based ones like Random Forest and GBMs that you want in your toolbox.
If you already know ML really well and constrained on time, you could skip ML videos. I have been doing some ML for a while and am actually getting lots of value from the ML class videos that Jeremy has posted.
If DL is completely new to you, I would suggest focus on DL and try the various things he wants us to try like running it on a new dataset from Kaggle or elsewhere. But eventually you do want to expand your Toolbox by learning some ML.
Hope this is useful.
5 Likes