Fix tests, use deeptime double well discrete data. (#1574)

Author: clonker

pyemma/coordinates/tests/test_nystroem_tica.py

@@ -75,7 +75,7 @@ class TestNystroemTICA_DoubleWell(unittest.TestCase):
     def setUpClass(cls):
         from pyemma.datasets import load_2well_discrete
         dw = load_2well_discrete()
-        v = dw.dtraj_T100K_dt10[:10000]
+        v = dw.dtraj[:10000]
         cls.T = v.size
         nstates = 100
         b = np.linspace(-1, 1, nstates)

pyemma/datasets/double_well_discrete.npz

@@ -20,21 +20,20 @@
 
 import numpy as np
 from deeptime.markov.msm import MarkovStateModel
+from deeptime.data import DoubleWellDiscrete
 
 from pyemma.msm import markov_model
 
 
-class DoubleWell_Discrete_Data(object):
+class DoubleWell_Discrete_Data(DoubleWellDiscrete):
     """ MCMC process in a symmetric double well potential, spatially discretized to 100 bins
 
     """
 
     def __init__(self):
-        from pkg_resources import resource_filename
-        filename = resource_filename('pyemma.datasets', 'double_well_discrete.npz')
-        datafile = np.load(filename)
-        self._dtraj_T100K_dt10 = datafile['dtraj']
-        self._P = datafile['P']
+        super().__init__()
+        self._dtraj_T100K_dt10 = self.dtraj[:]
+        self._P = self.transition_matrix
         self._msm_dt = MarkovStateModel(self._P)
         self._msm = markov_model(self._P)
 
@@ -70,11 +69,6 @@ def dtraj_T100K_dt10_n(self, divides):
             disc[divides[i]:divides[i + 1]] = i + 1
         return disc[self.dtraj_T100K_dt10]
 
-    @property
-    def transition_matrix(self):
-        """ Exact transition matrix used to generate the data """
-        return self._P
-
     @property
     def msm(self):
         """ Returns an MSM object with the exact transition matrix """

pyemma/datasets/double_well_discrete.py

@@ -94,7 +94,7 @@ class TestAMMDoubleWell(_tmsm):
     @classmethod
     def setUpClass(cls):
         import pyemma.datasets
-        cls.dtraj = pyemma.datasets.load_2well_discrete().dtraj_T100K_dt10
+        cls.dtraj = pyemma.datasets.load_2well_discrete().dtraj
         cls.E_ = np.linspace(0.01, 2.*np.pi, 66).reshape(-1,1)**(0.5)
         cls.m = np.array([1.9])
         cls.w = np.array([2.0])
@@ -381,7 +381,7 @@ class TestCoreAMMDoubleWell(_ctmsm):
     def setUpClass(cls):
         import pyemma.datasets
         cls.core_set = [34, 65]
-        cls.dtraj = pyemma.datasets.load_2well_discrete().dtraj_T100K_dt10
+        cls.dtraj = pyemma.datasets.load_2well_discrete().dtraj
         cls.E_ = np.linspace(0.01, 2.*np.pi, 66).reshape(-1,1)**(0.5)
         cls.m = np.array([1.9])
         cls.w = np.array([2.0])

pyemma/msm/tests/test_amm.py

@@ -19,6 +19,7 @@
 import unittest
 import numpy as np
 from deeptime.markov.tools.analysis import is_transition_matrix, is_reversible
+from numpy.testing import assert_allclose
 
 from pyemma.msm import bayesian_hidden_markov_model
 
@@ -29,7 +30,7 @@ class TestBHMM(unittest.TestCase):
     def setUpClass(cls):
         # load observations
         import pyemma.datasets
-        obs = pyemma.datasets.load_2well_discrete().dtraj_T100K_dt10
+        obs = pyemma.datasets.load_2well_discrete().dtraj
 
         # hidden states
         cls.nstates = 2
@@ -331,9 +332,9 @@ def test_separate_states(self):
                   np.array([2, 2, 2, 2, 2, 2, 1, 1, 1, 1, 1, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 2]),]
         hmm_bayes = bayesian_hidden_markov_model(dtrajs, 3, lag=1, separate=[0], nsamples=100, store_hidden=True)
         # we expect zeros in all samples at the following indexes:
-        pobs_zeros = [[0, 1, 2, 2, 2], [0, 0, 1, 2, 3]]
-        for s in hmm_bayes.samples:
-            assert np.allclose(s.observation_probabilities[pobs_zeros], 0)
+        pobs_zeros = ((0, 1, 2, 2, 2), (0, 0, 1, 2, 3))
+        for i, s in enumerate(hmm_bayes.samples):
+            assert_allclose(s.observation_probabilities[pobs_zeros], 0, err_msg=f"{i}")
         for strajs in hmm_bayes.sampled_trajs:
             assert strajs[0][0] == 2
             assert strajs[0][6] == 2
@@ -342,7 +343,7 @@ def test_initialized_bhmm(self):
         import pyemma.datasets as d
         import pyemma.msm
 
-        obs = d.load_2well_discrete().dtraj_T100K_dt10
+        obs = d.load_2well_discrete().dtraj
 
         init_hmm = pyemma.msm.estimate_hidden_markov_model(obs, 2, 10)
         bay_hmm = pyemma.msm.estimators.BayesianHMSM(nstates=init_hmm.nstates, lag=init_hmm.lag,

pyemma/msm/tests/test_bayesian_hmsm.py

@@ -30,7 +30,7 @@ def setUpClass(cls):
         # load observations
         import pyemma.datasets
         data = pyemma.datasets.load_2well_discrete()
-        obs_micro = data.dtraj_T100K_dt10
+        obs_micro = data.dtraj
 
         # stationary distribution
         pi_micro = data.msm.stationary_distribution

pyemma/msm/tests/test_bayesian_msm.py

@@ -155,7 +155,7 @@ def tearDown(self):
                 assert hasattr(restored, 'has_errors')
 
     def test_ck_msm(self):
-        MLMSM = msm.estimate_markov_model([self.double_well_data.dtraj_T100K_dt10_n6good], 40)
+        MLMSM = msm.estimate_markov_model([self.double_well_data.dtraj_n6good], 40)
         self.ck = MLMSM.cktest(2, mlags=[0, 1, 10], n_jobs=1)
         estref = np.array([[[1., 0.],
                             [0., 1.]],
@@ -178,7 +178,7 @@ def test_ck_msm(self):
         assert self.ck.predictions_conf[1] is None
 
     def test_its_bmsm(self):
-        BMSM = msm.bayesian_markov_model([self.double_well_data.dtraj_T100K_dt10_n6good], 40, reversible=True)
+        BMSM = msm.bayesian_markov_model([self.double_well_data.dtraj_n6good], 40, reversible=True)
         # also ensure that reversible bit does not flip during cktest
         assert BMSM.reversible
         self.ck = BMSM.cktest(2, mlags=[0, 1, 10], n_jobs=1)
@@ -216,7 +216,7 @@ def test_its_bmsm(self):
         assert np.allclose(self.ck.predictions[1], predRref, rtol=0.1, atol=10.0)
 
     def test_its_hmsm(self):
-        MLHMM = msm.estimate_hidden_markov_model([self.double_well_data.dtraj_T100K_dt10_n6good], 2, 10)
+        MLHMM = msm.estimate_hidden_markov_model([self.double_well_data.dtraj_n6good], 2, 10)
         self.ck = MLHMM.cktest(mlags=[0, 1, 10], n_jobs=1)
         estref = np.array([[[1., 0.],
                             [0., 1.]],
@@ -239,7 +239,7 @@ def test_its_hmsm(self):
         assert self.ck.predictions_conf[1] is None
 
     def test_its_bhmm(self):
-        BHMM = msm.bayesian_hidden_markov_model([self.double_well_data.dtraj_T100K_dt10_n6good], 2, 10)
+        BHMM = msm.bayesian_hidden_markov_model([self.double_well_data.dtraj_n6good], 2, 10)
         self.ck = BHMM.cktest(mlags=[0, 1, 10], n_jobs=1)
         estref = np.array([[[1., 0.],
                             [0., 1.]],

pyemma/msm/tests/test_cktest.py

@@ -66,7 +66,7 @@ def setUpClass(cls):
         import pyemma.datasets
         cls.core_set = [34, 65]
 
-        cls.dtraj = pyemma.datasets.load_2well_discrete().dtraj_T100K_dt10
+        cls.dtraj = pyemma.datasets.load_2well_discrete().dtraj
         nu = 1.*np.bincount(cls.dtraj)[cls.core_set]
         cls.statdist = nu/nu.sum()
 
@@ -698,7 +698,7 @@ class TestCoreMSM(unittest.TestCase):
     @classmethod
     def setUpClass(cls):
         from pyemma import datasets
-        cls.dtraj = datasets.load_2well_discrete().dtraj_T100K_dt10
+        cls.dtraj = datasets.load_2well_discrete().dtraj
 
     def test_core(self):
         core_set = [15, 16, 17, 45, 46, 47]

pyemma/msm/tests/test_cmsm.py

@@ -35,7 +35,7 @@ class TestMLHMM(unittest.TestCase):
     def setUpClass(cls):
         # load observations
         import pyemma.datasets
-        obs = pyemma.datasets.load_2well_discrete().dtraj_T100K_dt10
+        obs = pyemma.datasets.load_2well_discrete().dtraj.copy()
         obs -= np.min(obs)  # remove empty states
 
         # hidden states
@@ -465,7 +465,7 @@ def test_separate_states(self):
                   np.array([2, 2, 2, 2, 2, 2, 1, 1, 1, 1, 1, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 2]),]
         hmm = msm.estimate_hidden_markov_model(dtrajs, 3, lag=1, separate=[0])
         # we expect zeros in all samples at the following indexes:
-        pobs_zeros = [[0, 1, 2, 2, 2], [0, 0, 1, 2, 3]]
+        pobs_zeros = ((0, 1, 2, 2, 2), (0, 0, 1, 2, 3))
         assert np.allclose(hmm.observation_probabilities[pobs_zeros], 0)
 
 if __name__=="__main__":

pyemma/msm/tests/test_hmsm.py

@@ -303,7 +303,7 @@ def test_its_cmsm_defined_core(self):
         core_set = [0, 5]
         estimator = msm.estimators.MaximumLikelihoodMSM(core_set=core_set)
         its = msm.ImpliedTimescales(estimator, lags=[1, 10, 100, 1000], n_jobs=1)
-        its.estimate([self.double_well_data.dtraj_T100K_dt10_n6good])
+        its.estimate([self.double_well_data.dtraj_n6good])
         assert its.models[0].n_cores == 2
         ref = np.array([[339.22244263],
                         [334.56862305],