I was trying to deploy a new version of my app to Heroku with conda-buildpack for numpy and scipy. However, an extra mkl-11.3.1 was downloaded and installed which is 121.2 MB.
Fetching package metadata: ....
Solving package specifications: ..................
Package plan for installation in environment /app/.heroku/miniconda:
The following packages will be downloaded:
package | build
---------------------------|-----------------
mkl-11.3.1 | 0 121.2 MB
numpy-1.10.4 | py27_0 6.0 MB
scipy-0.17.0 | np110py27_1 29.6 MB
------------------------------------------------------------
Total: 156.8 MB
The following NEW packages will be INSTALLED:
mkl: 11.3.1-0
The following packages will be UPDATED:
numpy: 1.10.2-py27_0 --> 1.10.4-py27_0
scipy: 0.16.1-np110py27_0 --> 0.17.0-np110py27_1
Anyway, nothing was changed in my conda-requirements.txt file which is
numpy
scipy
Anyone knows what's going on? Thank you so much!
Conda started compiling numpy/scipy/numexpr with the Intel MKL as of v2.5: https://www.continuum.io/blog/developer-blog/anaconda-25-release-now-mkl-optimizations
If you add nomkl to the beginning of your conda-requirements.txt file, it will use the non-mkl optimized binaries, and won't download the mkl package.