Possible bug: dare_test_tjm_2 and dare_test_tjm_3 with accelerate/mkl #84

albop · 2014-10-16T16:52:26Z

On my laptop, I got the following output when running the tests this morning:

======================================================================
ERROR: quantecon.tests.test_ricatti.dare_test_tjm_2
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/pablo/.local/opt/anaconda/lib/python2.7/site-packages/nose/case.py", line 197, in runTest
    self.test(*self.arg)
  File "/home/pablo/Programming/quantecon/quant-econ/quantecon/tests/test_ricatti.py", line 50, in dare_test_tjm_2
    X = solve_discrete_riccati(A, B, Q, R)
  File "/home/pablo/Programming/quantecon/quant-econ/quantecon/matrix_eqn.py", line 162, in solve_discrete_riccati
    Q_tilde = - Q + dot(C.T, solve(Z, C + gamma * BTA)) + gamma * I
  File "/home/pablo/.local/opt/anaconda/lib/python2.7/site-packages/numpy/linalg/linalg.py", line 381, in solve
    r = gufunc(a, b, signature=signature, extobj=extobj)
  File "/home/pablo/.local/opt/anaconda/lib/python2.7/site-packages/numpy/linalg/linalg.py", line 90, in _raise_linalgerror_singular
    raise LinAlgError("Singular matrix")
LinAlgError: Singular matrix

======================================================================
ERROR: quantecon.tests.test_ricatti.dare_test_tjm_3
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/pablo/.local/opt/anaconda/lib/python2.7/site-packages/nose/case.py", line 197, in runTest
    self.test(*self.arg)
  File "/home/pablo/Programming/quantecon/quant-econ/quantecon/tests/test_ricatti.py", line 66, in dare_test_tjm_3
    X = solve_discrete_riccati(A, B, Q, R)
  File "/home/pablo/Programming/quantecon/quant-econ/quantecon/matrix_eqn.py", line 162, in solve_discrete_riccati
    Q_tilde = - Q + dot(C.T, solve(Z, C + gamma * BTA)) + gamma * I
  File "/home/pablo/.local/opt/anaconda/lib/python2.7/site-packages/numpy/linalg/linalg.py", line 381, in solve
    r = gufunc(a, b, signature=signature, extobj=extobj)
  File "/home/pablo/.local/opt/anaconda/lib/python2.7/site-packages/numpy/linalg/linalg.py", line 90, in _raise_linalgerror_singular
    raise LinAlgError("Singular matrix")
LinAlgError: Singular matrix

The problem disappeared when I upgraded Anaconda, and removed the MKL and accelerate packages (since I don't have a license anymore):

The following packages will be REMOVED:

    accelerate:   1.5.1-np18py27_p0 
    mkl:          11.1-np18py27_p2  

The following packages will be UPDATED:

    numexpr:      2.3.1-np19py27_p0  [mkl] --> 2.3.1-np19py27_0 
    numpy:        1.9.0-py27_p0      [mkl] --> 1.9.0-py27_0     
    scikit-learn: 0.15.2-np19py27_p0 [mkl] --> 0.15.2-np19py27_0
    scipy:        0.14.0-np19py27_p0 [mkl] --> 0.14.0-np19py27_0

Has anybody run the tests with the accelerate/mkl packages ?

The text was updated successfully, but these errors were encountered:

cc7768 · 2014-10-17T14:27:23Z

@albop I replicated this when I ran tests with Python 3.4 using MKL numpy through the Anaconda accelerate package.

It looks like the following is happening:

In line 160 of matrix_eqn.py we are checking the condition of a matrix Z to make sure that we can invert it in line 162. In both dare_test_tjm_2 and dare_test_tjm_3 (the tests that fail) the condition of Z under gamma = 0. is on the magnitude of 1e17, which seems to suggest that it is likely ill-conditioned. @jstac Should we change this to something like cn < BIGNUMBER?

I might look into it a little more later, I just wanted to at least figure out what was happening before I forgot about it.

cc7768 · 2014-10-17T14:30:40Z

The reason we don't get this problem when we aren't on MKL is because the default linear algebra pacakge is LAPACK and it finds that the condition of the matrices is inf which is what we are checking for.

sglyon · 2015-01-28T21:50:24Z

Is still still an issue with the re-vamped matrix tools functions?

cc7768 · 2015-01-28T21:54:25Z

We have done nothing to address these issues for now. It has to do with LAPACK and MKL being slightly different.

oyamad · 2016-02-18T13:45:39Z

The same tests failed recently: https://travis-ci.org/QuantEcon/QuantEcon.py/builds/109609660.
Is it because Numpy and related packages now depend on MKL? See https://www.continuum.io/blog/developer-blog/anaconda-25-release-now-mkl-optimizations.

mmcky · 2016-02-24T21:28:37Z

This is now high priority due to tests failing under the latest anaconda distribution. Is anyone aware of a root cause of this issue before I dive in?

I cannot seem to replicate this on a fresh py35 environment. Is it possible this has been fixed upstream in either numpy or by continuum?

mmcky · 2016-02-25T15:24:23Z

I have updated the TravisCI environment. It seems to pass all tests suggesting an upstream fix. If someone had some time to run nosetests over the QuantEcon.py package after updating conda to the latest version - that would be appreciated.

Then I propose we close the issue.

jstac · 2016-02-25T15:39:11Z

I guess this is a separate issue but here's my output

======================================================================
ERROR: Failure: ImportError (The code previously contained in the quantecon.models subpackage has been migrated to the QuantEcon.applications (https://github.com/QuantEcon/QuantEcon.applications) repo)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/john/anaconda3/lib/python3.5/site-packages/nose/failure.py", line 39, in runTest
    raise self.exc_val.with_traceback(self.tb)
  File "/home/john/anaconda3/lib/python3.5/site-packages/nose/loader.py", line 418, in loadTestsFromName
    addr.filename, addr.module)
  File "/home/john/anaconda3/lib/python3.5/site-packages/nose/importer.py", line 47, in importFromPath
    return self.importFromDir(dir_path, fqname)
  File "/home/john/anaconda3/lib/python3.5/site-packages/nose/importer.py", line 94, in importFromDir
    mod = load_module(part_fqname, fh, filename, desc)
  File "/home/john/anaconda3/lib/python3.5/imp.py", line 244, in load_module
    return load_package(name, filename)
  File "/home/john/anaconda3/lib/python3.5/imp.py", line 216, in load_package
    return _load(spec)
  File "<frozen importlib._bootstrap>", line 693, in _load
  File "<frozen importlib._bootstrap>", line 673, in _load_unlocked
  File "<frozen importlib._bootstrap_external>", line 662, in exec_module
  File "<frozen importlib._bootstrap>", line 222, in _call_with_frames_removed
  File "/home/john/sync_dir/books/quant-econ/QuantEcon.py/quantecon/models/__init__.py", line 1, in <module>
    raise ImportError("The code previously contained in the quantecon.models subpackage has been migrated to the QuantEcon.applications (https://github.com/QuantEcon/QuantEcon.applications) repo")
ImportError: The code previously contained in the quantecon.models subpackage has been migrated to the QuantEcon.applications (https://github.com/QuantEcon/QuantEcon.applications) repo

----------------------------------------------------------------------
Ran 263 tests in 31.321s

FAILED (SKIP=1, errors=1)
Closing remaining open files:/home/john/sync_dir/books/quant-econ/QuantEcon.py/quantecon/tests/data/testing_data.h5...done

mmcky · 2016-02-25T15:45:05Z

Thanks @jstac this is actually an error that results in the notice for removing the models subpackage. So when models loads it throws the ImportError.

I should have been more specific with nosetests, Travis runs the following:

nosetests --exclude=models

jstac · 2016-02-25T15:48:49Z

With latest anaconda:

nosetests --exclude=models                                +1938 10:47 ❰─┘
.........................................................................................................................S.............................................................................................................................................
----------------------------------------------------------------------
Ran 262 tests in 31.575s

OK (SKIP=1)
Closing remaining open files:/home/john/sync_dir/books/quant-econ/QuantEcon.py/quantecon/tests/data/testing_data.h5...done

mmcky · 2016-02-25T20:34:20Z

That looks like the test suite passes on your machine also.
I plan to close this issue tomorrow - unless anyone objects.

oyamad · 2017-10-23T01:24:09Z

I got the same errors today.

anaconda                  5.0.0            py36hd9bc8a5_0

mkl                       2018.0.0             hc285769_4

numpy                     1.13.1           py36h93d791d_2

mmcky · 2017-10-23T01:44:50Z

Hey @oyamad I wasn't able to replicate this test failure.
I created the latest anaconda environment with

mkl                       2018.0.0             hb491cac_4 
anaconda                  5.0.0.1          py36hfb0b618_1
numpy                     1.13.1           py36h5bc529a_2

oyamad · 2017-10-23T01:48:41Z

I am on Mac:

$ python
Python 3.6.2 |Anaconda, Inc.| (default, Sep 21 2017, 18:29:43) 
[GCC 4.2.1 Compatible Clang 4.0.1 (tags/RELEASE_401/final)] on darwin

mmcky · 2017-10-23T01:50:41Z

OK thanks for reporting @oyamad. Maybe it's time to add CircleCI and Appveyor here. We have these on conda-forge but that isn't run as frequently as this master repo.

oyamad · 2017-10-23T02:29:44Z

Re #84 (comment), on my machine I have cn = 3.591469197922365e+17 for gamma = 0. (for both dare_test_tjm_2 and dare_test_tjm_3).

cc7768 added the tests label May 7, 2015

mmcky added the bug label Feb 24, 2016

mmcky closed this as completed Mar 15, 2016

oyamad reopened this Oct 23, 2017

oyamad mentioned this issue Oct 23, 2017

TEST: Set atol #357

Merged

This was referenced Oct 23, 2017

qe.solve_discrete_riccati versus scipy.linalg.solve_discrete_are #360

Closed

FIX: Replace np.isfinite(cn) with cn * EPS < 1 in solve_discrete_riccati #361

Merged

mmcky closed this as completed in #361 Oct 23, 2017

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Possible bug: dare_test_tjm_2 and dare_test_tjm_3 with accelerate/mkl #84

Possible bug: dare_test_tjm_2 and dare_test_tjm_3 with accelerate/mkl #84

albop commented Oct 16, 2014

cc7768 commented Oct 17, 2014

cc7768 commented Oct 17, 2014

sglyon commented Jan 28, 2015

cc7768 commented Jan 28, 2015

oyamad commented Feb 18, 2016

mmcky commented Feb 24, 2016

mmcky commented Feb 25, 2016

jstac commented Feb 25, 2016

mmcky commented Feb 25, 2016

jstac commented Feb 25, 2016

mmcky commented Feb 25, 2016

oyamad commented Oct 23, 2017

mmcky commented Oct 23, 2017

oyamad commented Oct 23, 2017

mmcky commented Oct 23, 2017

oyamad commented Oct 23, 2017

Possible bug: dare_test_tjm_2 and dare_test_tjm_3 with accelerate/mkl #84

Possible bug: dare_test_tjm_2 and dare_test_tjm_3 with accelerate/mkl #84

Comments

albop commented Oct 16, 2014

cc7768 commented Oct 17, 2014

cc7768 commented Oct 17, 2014

sglyon commented Jan 28, 2015

cc7768 commented Jan 28, 2015

oyamad commented Feb 18, 2016

mmcky commented Feb 24, 2016

mmcky commented Feb 25, 2016

jstac commented Feb 25, 2016

mmcky commented Feb 25, 2016

jstac commented Feb 25, 2016

mmcky commented Feb 25, 2016

oyamad commented Oct 23, 2017

mmcky commented Oct 23, 2017

oyamad commented Oct 23, 2017

mmcky commented Oct 23, 2017

oyamad commented Oct 23, 2017