numpy.disutils.system_info.NotFoundError: no lapack/blas resources found

Question

Problem: Linking numpy to correct Linear Algebra libraries. Process is so complicated that I might be looking for the solution 6th time and I have no idea whats going wrong. I am on Ubuntu 12.04.5. I reinstalled blas and lapack and then reinstalled numpy using pip. I did that in system environment and then also tried in virtualenv environment. None of the things seem to be working.

Here is my numpy.__config__.show():

lapack_info:
  NOT AVAILABLE
lapack_opt_info:
  NOT AVAILABLE
openblas_lapack_info:
  NOT AVAILABLE
blas_info:
  NOT AVAILABLE
atlas_3_10_blas_threads_info:
  NOT AVAILABLE
atlas_threads_info:
  NOT AVAILABLE
blas_src_info:
  NOT AVAILABLE
atlas_3_10_threads_info:
  NOT AVAILABLE
atlas_blas_info:
  NOT AVAILABLE
atlas_3_10_blas_info:
  NOT AVAILABLE
lapack_src_info:
  NOT AVAILABLE
atlas_blas_threads_info:
  NOT AVAILABLE
openblas_info:
  NOT AVAILABLE
blas_mkl_info:
  NOT AVAILABLE
blas_opt_info:
  NOT AVAILABLE
atlas_info:
  NOT AVAILABLE
atlas_3_10_info:
  NOT AVAILABLE
lapack_mkl_info:
  NOT AVAILABLE
mkl_info:
  NOT AVAILABLE

This is a very common error and has lots of solution depending on the user's environment and history. I recently went through the whole process of installing BLAS and LAPACK as instructed here: https://stackoverflow.com/a/9173550/3413239. I was able to successfully install both.

However, I initially had these libraries: (ls /usr/lib | grep blas)

libopenblas.so.0
openblas-base/
libblas.so.3gf
libblas.3

How should I link my numpy to linear algebra libraries. Is it mandatory to have all the linaear algebra libraries? i.e. lapack, blas, atlas?

I am assuming once the numpy is linked all other modules like scipy and Orange will work.

Solution 1

It's possible that you have unmet dependencies. When I had this error, I was able to fix it by installing two packages:

sudo apt-get install libblas-dev  liblapack-dev

Solution 2

Install following libraries (Ubuntu 20.04):

sudo apt-get install libatlas-base-dev libblas3 liblapack3 liblapack-dev libblas-dev gfortran

and then install python packages using pip.

Solution 3

If anyone has a similar issue with Scipy installation on FreeBSD - it can be fixed quite simply by installing necessary packages on the system itself using pkg command:

pkg install blas
pkg install lapack

Comments

[ChickenFeet]

I understand this question is Ubuntu specific, but would you happen to know how to solve this same issue with Windows?