ADVI Is Much Slower than that Showed in the tutorial case

I am trying to repeat Bayesian neural network (i.e: a simple binary classification problem that’s not linearly separable) from the PyMC3 tutorial case “Variational Inference: Bayesian Neural Networks”. However, it is much slower than that showed in the case (speed up to 3139.84it/s). The code and data used were the same with the case.
Actual result:

PC: CPU—— Core(TM) I7-4720HQ ; RAM——16G.

I would like to find out how to speed it up ? Thank you for help~

How did you install PyMC3?
Did you see any warning when you import pymc3 and/or theano? My guess is that your BLAS library is not set up properly

I installed by “pip install pymc3” in Anaconda.
When import theano, it shows:

Maybe your guess is right. But how to set up the library properly? :confused:

python version: 3.6.3 OS: windows7

The most painless way is to install Anaconda and use conda to configure your environment and instead related package.
Otherwise, you can check

1 Like

I have solved the problem by “conda install mingw libpython numpy scipy mkl-service libpython m2w64-toolchain”. Thank junpenglao :blush: