Similar question as here but now on Python packages. Currently, the CVXPY is missing in Azure ML. I am also trying to get other solvers such as GLPK, CLP and COINMP working in Azure ML.
How can I install Python packages in Azure ML?
Update about trying to install the Python packages not found in Azure ML.
I did as instructed by Peter Pan but I think the 32bits CVXPY files are wrong for the Anaconda 4 and Python 3.5 in Azure ML, logs and errors are here.
[Information] Running with Python 3.5.1 |Anaconda 4.0.0 (64-bit)| (default, Feb 16 2016, 09:49:46) [MSC v.1900 64 bit (AMD64)]
Update 2 with win_amd64 files (paste here)
[Information] Extracting Script Bundle.zip to .\Script Bundle [Information] File Name Modified Size [Information] cvxopt-1.1.9-cp35-cp35m-win_amd64.whl 2017-06-07 01:03:34 1972074 [Information] __MACOSX/ 2017-06-07 01:26:28 0 [Information] __MACOSX/._cvxopt-1.1.9-cp35-cp35m-win_amd64.whl 2017-06-07 01:03:34 452 [Information] cvxpy-0.4.10-py3-none-any.whl 2017-06-07 00:25:36 300880 [Information] __MACOSX/._cvxpy-0.4.10-py3-none-any.whl 2017-06-07 00:25:36 444 [Information] ecos-2.0.4-cp35-cp35m-win_amd64.whl 2017-06-07 01:03:40 56522 [Information] __MACOSX/._ecos-2.0.4-cp35-cp35m-win_amd64.whl 2017-06-07 01:03:40 450 [Information] numpy-1.13.0rc2+mkl-cp35-cp35m-win_amd64.whl 2017-06-07 01:25:02 127909457 [Information] __MACOSX/._numpy-1.13.0rc2+mkl-cp35-cp35m-win_amd64.whl 2017-06-07 01:25:02 459 [Information] scipy-0.19.0-cp35-cp35m-win_amd64.whl 2017-06-07 01:05:12 12178932 [Information] __MACOSX/._scipy-0.19.0-cp35-cp35m-win_amd64.whl 2017-06-07 01:05:12 452 [Information] scs-1.2.6-cp35-cp35m-win_amd64.whl 2017-06-07 01:03:34 78653 [Information] __MACOSX/._scs-1.2.6-cp35-cp35m-win_amd64.whl 2017-06-07 01:03:34 449 [Information] [ READING ] 0:00:00 [Information] Input pandas.DataFrame #1: [Information] Empty DataFrame [Information] Columns: [1] [Information] Index: [] [Information] [ EXECUTING ] 0:00:00 [Information] [ WRITING ] 0:00:00
where
import cvxpy
,import cvxpy-0.4.10-py3-none-any.whl
orcvxpy-0.4.10-py3-none-any
do not work soHow can I use the following wheel files downloaded from here to use the external Python packages not found in Azure ML?
Update about permission problem about importing cvxpy (paste here)
[Error] ImportError: No module named 'canonInterface'
where the ZIP Bundle is organised a bit differently, the content of each wheel downloaded to a folder and the content having all zipped as a ZIP Bundle.
According to the
Limitations
andTechnical Notes
ofExecute Python Script
tutorial, the only way to add custom Python modules is via the zip file mechanism to package the modules and all dependencies.For example to install
CVXPY
, as below.CVXPY
and its dependencies likeCVXOPT
.cvxpy
andcvxopt
, etc as a zipped file with your script.If you were using IPython, you also can try to install the Python Package via the code
!pip install cvxpy
.And there are some similar SO threads which may be helpful for you, as below.
Hope it helps.
Update:
For IPython interface of Azure ML, you move to the
NOTEBOOKS
tab to create a notebook viaADD TO PROJECT
button at the bottom of the page, as the figure below.Or you can directly login to the website
https://notebooks.azure.com
to use it.I will demo Peter Pan's answer here and I show first a working experiment with external package. Building the ZIP Bundle for the external package CVXPY was not that straight-forward but educational.
Demo about external package using ZIP Bundle in Azure ML
This is a working example about using an additional package in Python Script module provided by Smit Shah
where
so the ZIP bundle has been done in a specific way such that packages' dist-info folders are removed and only the folders having the exact name of the package is included
by which I was able to run
import cvxpy
, notice that new packagesmultiplocess
andCVXCanon
are included. Yet in solving a problem, I got error withecos
module missing next. So I created a new directoryecos
for all files found in the originalecos.*.wheel
package so the updated ZIP Bundle looks like thisby which I was able to solve the following convex optimisation problem with DCP tool called CVXPY.
Code inside Python Script module that solves an optimisation problem with CVXPY in Azure ML
Further reading