I was trying to install pymc3 with Miniconda in Windows. The installation was certainly non-trivial and took me many hours of trouble-shooting to understand what is going on.
I have read many posts from different people on this question and found this guide to be quite helpful:
pymc3 installation guide in Windows
After following through the steps recommended here, I still came across the warning when importing pymc3 as pm in a jupyter-notebook:
WARNING (theano.tensor.blas): Using NumPy C-API based implementation for BLAS functions.
Below are some solutions that I have tested and did not work to remove this warning:
conda install -c conda forge libpython blas mkl-service m2w64-toolchain in addition to the packages bundled with pymc3 (theano-pymc3 and arviz) in the first conda environment.
Downloaded GCC compiler and make sure it is on the global path in the correct order as suggested by the first guideline and the official guideline Official pymc3 installation in the second conda environment and installed without m2w64-toolchain.
I have also tried using the GCC compiler from Rtools and updated the path in the environment variable in the 3rd conda environment.
I think I have exhausted all the different options I have read and still didn't manage to remove this warning. Is there anything that I have missed out in trouble-shooting this warning?
I understand that this warning doesn't stop the computation, and would only affect the computational performance. Because I have quite a large dataset, I would certainly appreciate a feasible solution to fix this problem :)
Thank you for any feedback!
According to https://zhuanlan.zhihu.com/p/394033647
to create .theanorc.txt
may solve this problem
Related
I remember that g++ 9.3.0 required to link against libtbb to use parallel STL algorithms, otherwise it gave compilation errors. I have a docker container using an ubuntu image where I have installed g++-10 only and I can use the algorithms without writing -ltbb. How can I explain this? I searched everywhere and there is no libtbb anywhere, so I am assuming g++-10 no longer needs it? Where can I read some documentation about this fact that it does not need it or is required up to version x.x.x? Thanks for clarifying.
I'm aware that this question has been asked before, but after doing quite a bit of research, I find myself stuck with my problem.
I'm currently comfortable in R, and would like to gradually learn Python basics. I've installed python from the official website, and opted for the Spyder IDE. I ran a few basic Python commands such a printing hello. So far so good. Now I'd like to learn by working with datasets.
My understanding is that you need the Pandas package for that (or it is at least recommended). And here start the many, many issues. I read that you need pip to install packages. To my surprise, I had to install that through the windows command prompt. After many failed attempts, I managed, and later even managed installing the packages. However, when I go back to the Spyder IDE and try to run the generic commands, it always returns invalid syntax to me.
I'm happy to share screenshots, or more details if the problem isn't clear enough. I expected to find a steep learning curve at the very beginning, but not to struggle with packages.
Any help greatly appreciated
I am I trying to install SCIP-SDP. As I am not a Linux user, I have found the instructions for installing SCIP very confusing, and have not managed to install it.
I have then chosen to use SCIP through Matlab by using Opti. I have managed to solve some LPs with SCIP through Opti. Is it possible to use SCIP-SDP through Opti? If so, could you please give me some guidelines on how to do it?
There exists an interface from the optitoolbox to SCIP-SDP, but this is quite old and will only work with SCIP-SDP 1.0, but since I haven't used the toolbox myself, I can't give you any more details. There is also the possibility to use SCIP-SDP via the neos-server, but this is also only SCIP-SDP 2.0 (and you should give a settings file, since the default settings are sometimes a little bit strange). For the most current version, you will unfortunately need to install it through the console, and this is really only tested on Linux. Perhaps you could try to use a virtual machine? If there are any specific problems with installing SCIP or SCIP-SDP (preferably on Linux), then feel free to ask, either here or via mail.
i want to work with NumPy. Now i'm trying to install NumPy on Windows. I do not understand, what i have to do to get it working. What does "building" mean? is there any difference between the installation and "building from source".
Thanks for any help.
Probably the easiest approach to installing Numpy (and the rest of the Python 'scientific stack' including SciPy an many others) is to use Continuum Analytics' "Anaconda" tools.
http://docs.continuum.io/anaconda/install#windows-install
Following these docs should get you everything you need.
I am trying to link NumPy and SciPy (under Python 3, under Fedora 20), using only RPM packages (so, no pip, manual builds, etc).
I've tried various stuff, but
import numpy as np
import scipy
scipy.__config__.show()
np.__config__.show()
always gives me the usual "I'm using ATLAS" info.
I've tried playing with alternatives, as suggested here, here, and in several other resources, but alternatives --list | grep blas is always empty. I've tried adding a libblas.so.3 alternative, linking to libopenblas.so, but the behaviour of the above test didn't change.
I've also tried to find out where and how SciPy/NumPy load these libraries, but couldn't find it.
I don't have _dotblas.so (often mentioned in various instructions related to this); however, I do have _dotblas.cpython-33m.so, cblas.cpython-33m.so, and fblas.cpython-33m.so, but none of them links to libblas.so.3 (as suggested in the previous link).
Removing ATLAS is not an option, as it drags many more packages with it (Gimp, for example).
I've also installed a fresh Fedora 20 inside a virtual machine. NumPy/SciPy from yum insisted on ATLAS. When I removed those and installed from pip, they had no problem using OpenBLAS (but then I couldn't easily switch ATLAS). I want both, with an easy switch, so I can perform various tests to compare how these two perform, and (re)installing between yum and pip is obviously not a solution.
I want to avoid rebuilding and manual installation, for easier future updates and because I need this on several somewhat different machines. It doesn't seem to me like it's a weird thing to want to do, telling these modules to load one library or another, without major reinstalls whenever I want to make the change.