Repository navigation
MKL FATAL ERROR on Google Colab with mkl==2019.0 #730
Description
Activity
Hi,
python-control does not call any BLAS/LAPACK library routine directly. We don't have and won't add any requirement on MKL, OpenBLAS or any other implemetation of BLAS/LAPACK directly to python-control.
This is most likely an issue with the MKL/NumPy/Slycot combination of your instance in Google Colab. And as you have found out, it seems to be fixed by updating to a more recent version. I am not familiar with the innards of Colab, but it seems that
!pip install slycotis the culprit here. On a regular Python environment this would download the source and compile the Fortran routines with the headers present on the system.BTW, numpy on Colab says, that is not using MKL at all:
import numpy as np np.show_config()``
blas_mkl_info: NOT AVAILABLE blis_info: NOT AVAILABLE openblas_info: libraries = ['openblas', 'openblas'] library_dirs = ['/usr/local/lib'] language = c define_macros = [('HAVE_CBLAS', None)] runtime_library_dirs = ['/usr/local/lib'] blas_opt_info: libraries = ['openblas', 'openblas'] library_dirs = ['/usr/local/lib'] language = c define_macros = [('HAVE_CBLAS', None)] runtime_library_dirs = ['/usr/local/lib'] lapack_mkl_info: NOT AVAILABLE openblas_lapack_info: libraries = ['openblas', 'openblas'] library_dirs = ['/usr/local/lib'] language = c define_macros = [('HAVE_CBLAS', None)] runtime_library_dirs = ['/usr/local/lib'] lapack_opt_info: libraries = ['openblas', 'openblas'] library_dirs = ['/usr/local/lib'] language = c define_macros = [('HAVE_CBLAS', None)] runtime_library_dirs = ['/usr/local/lib'] Supported SIMD extensions in this NumPy install: baseline = SSE,SSE2,SSE3 found = SSSE3,SSE41,POPCNT,SSE42,AVX,F16C,FMA3,AVX2 not found = AVX512F,AVX512CD,AVX512_KNL,AVX512_KNM,AVX512_SKX,AVX512_CLX,AVX512_CNL,AVX512_ICLIf you compile Slycot with
-DBLA_VENDOR=Generic, then you won't have to update MKL:%env SKBUILD_CONFIGURE_OPTIONS="-DBLA_VENDOR=Generic" !pip install slycot control import control as ct
env: SKBUILD_CONFIGURE_OPTIONS="-DBLA_VENDOR=Generic" Collecting slycot Downloading slycot-0.4.0.0.tar.gz (1.5 MB) |████████████████████████████████| 1.5 MB 8.8 MB/s Installing build dependencies ... done Getting requirements to build wheel ... done Preparing wheel metadata ... done Collecting control Downloading control-0.9.1.tar.gz (357 kB) |████████████████████████████████| 357 kB 39.9 MB/s Requirement already satisfied: numpy in /usr/local/lib/python3.7/dist-packages (from slycot) (1.21.6) Requirement already satisfied: scipy in /usr/local/lib/python3.7/dist-packages (from control) (1.4.1) Requirement already satisfied: matplotlib in /usr/local/lib/python3.7/dist-packages (from control) (3.2.2) Requirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->control) (3.0.8) Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.7/dist-packages (from matplotlib->control) (0.11.0) Requirement already satisfied: python-dateutil>=2.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->control) (2.8.2) Requirement already satisfied: kiwisolver>=1.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->control) (1.4.2) Requirement already satisfied: typing-extensions in /usr/local/lib/python3.7/dist-packages (from kiwisolver>=1.0.1->matplotlib->control) (4.2.0) Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.7/dist-packages (from python-dateutil>=2.1->matplotlib->control) (1.15.0) Building wheels for collected packages: slycot, control Building wheel for slycot (PEP 517) ... done Created wheel for slycot: filename=slycot-0.4.0-cp37-cp37m-linux_x86_64.whl size=1418275 sha256=99f50c169426a9233c293bd29c813c51e4d9abb480661c45964ae2885fe47a2b Stored in directory: /root/.cache/pip/wheels/fa/b2/19/1e6568896fbabaf2a03d2fb6575211d281afcb9e00ce026cf6 Building wheel for control (setup.py) ... done Created wheel for control: filename=control-0.9.1-py2.py3-none-any.whl size=364703 sha256=294faf3306345fc5a496a380203c8378d79344075a7b0ae67358db5f613cbea0 Stored in directory: /root/.cache/pip/wheels/b1/50/22/8005107713828997d1af664fbd629c7f935a1b87fc2f4ba6c1 Successfully built slycot control Installing collected packages: slycot, control Successfully installed control-0.9.1 slycot-0.4.0ct.tf2ss(ct.tf([1],[1,1]))
\[ \left(\begin{array}{rll|rll} -1\phantom{.}&\hspace{-1em}&\hspace{-1em}\phantom{\cdot}&1\phantom{.}&\hspace{-1em}&\hspace{-1em}\phantom{\cdot}\\ \hline 1\phantom{.}&\hspace{-1em}&\hspace{-1em}\phantom{\cdot}&0\phantom{.}&\hspace{-1em}&\hspace{-1em}\phantom{\cdot}\\ \end{array}\right) \]Now you have to figure out how to get Colab to render LaTeX output like a real Jupyter Notebook or ignore it and use the regular
__repr__()Hi @bnavigator / Ben!
I saw your comments here and decided to get in touch. I am also trying to run a very simple code on Google Colab using the Control + slycot library, but the Kernel restarts (crashes).
Were you able to solve the problem?
Best regards,
Nicolas Spogis
Colab Errors:
WARNING:root:kernel f5072b3f-42c0-4442-bb37-dd67744b63be restarted
KernelRestarter: restarting kernel (1/5), keep random ports
Intel MKL FATAL ERROR: Cannot load libmkl_avx2.so or libmkl_def.so.
INTEL MKL ERROR: /lib/x86_64-linux-gnu/libmkl_avx2.so: undefined symbol: mkl_sparse_optimize_bsr_trsm_i8.
Code on Google Colab:
!apt-get install gfortran cmake --fix-missing
!apt-get install libblas-dev liblapack-devimport os
os.environ['BLA_VENDOR']="Generic"
!echo Using $$BLA_VENDOR BLAS implementation!pip install -v slycot
!pip install control
import control
import slycot
import matplotlib.pyplot as pltDefine transfer function of the system
s = control.tf('s')
K1 = 5.0
T1 = 5.0
P1 = K1/(T1*s+1)K2 = 10.0
T2 = 20.0
P2 = K2/(T2*s+1)K3 = 5.0
T3 = 5.0
P3 = K3/(T3*s+1)K4 = 5.0
T4 = 5.0
P4 = K4/(T4*s+1)G = control.append(P1, P2, P3, P4) #Google Colab Crash HERE!
Now not crash, but i have this error:
AttributeError Traceback (most recent call last) [<ipython-input-11-beb8f34beafd>](https://localhost:8080/#) in <cell line: 1>() ----> 1 G = control.append(P1, P2, P3, P4) 2 frames [/usr/local/lib/python3.10/dist-packages/slycot/transform.py](https://localhost:8080/#) in td04ad(rowcol, m, p, index, dcoeff, ucoeff, tol, ldwork) 622 *(dcoeff.shape + expectedshape)), 623 -5) --> 624 out = _wrapper.td04ad_c(m,p,index,dcoeff,ucoeff,n,tol,ldwork) 625 else: 626 raise SlycotParameterError("Parameter rowcol had an illegal value", -1) AttributeError: module 'slycot._wrapper' has no attribute 'td04ad_c'This is my Google Colab: https://colab.research.google.com/drive/1hb9XDIVeMirQ2gevnaU282yFTs4-KS_t?usp=sharing
I'm already getting a numpy error while importing control when running your colab.
My own colab notebook still works, including your code:
https://colab.research.google.com/drive/1kKaK3jI_omOTrLCLQGzSs1uTowN5Cx_v?usp=sharing
Hi @bnavigator / Ben!
Thank you very very very much!
You help me a lot.
As "The Cure" said! It's Friday, I'm in love!
Best regards,
Nicolas
When using
controlversion0.9.1on Google Colab andmklversion is2019.0the session crashes when runningts2ss()with the following runtime log:Steps to recreate:
!pip install control!pip install slycot(likely not related)!pip show mkl(verify mkl version, my sessions have all been2019.0so far)Session crashes with previous error in runtime logs.
The fix I have found is to update the
mklversion to2021.4.0. I am not familiar enough with the version changes nor have I tried othermklversions, so I cannot say what other versions work.!pip install mkl==2021.4.0Probably should add a requirement on
mklversion when installingcontrolin Colab.I have not tried to reproduce this on a local Jupyter notebook.