Minor issue with newest numpy version

[pitmanme]

PyEMMA appears to not be compatible with the most recent NumPy version. I detail the version related bug I ran into below and a quick fix I implemented on my installation of PyEMMA so that I didn't have to backtrack my NumPy installation.

Here is the Traceback :

  File "/Users/mpitman/work/rfe/pca_background/pca/tica.py", line 33, in <module>
    tica_obj = coor.tica(inp, lag=lag, dim=2, kinetic_map=False)
  File "/opt/miniconda3/envs/mdtools/lib/python3.9/site-packages/decorator.py", line 232, in fun
    return caller(func, *(extras + args), **kw)
  File "/opt/miniconda3/envs/mdtools/lib/python3.9/site-packages/pyemma/util/annotators.py", line 218, in _deprecated
    return func(*args, **kw)
  File "/opt/miniconda3/envs/mdtools/lib/python3.9/site-packages/pyemma/coordinates/api.py", line 1287, in tica
    res.estimate(data, chunksize=cs)
  File "/opt/miniconda3/envs/mdtools/lib/python3.9/site-packages/pyemma/coordinates/transform/tica.py", line 163, in estimate
    return super(TICA, self).estimate(X, **kwargs)
  File "/opt/miniconda3/envs/mdtools/lib/python3.9/site-packages/pyemma/coordinates/data/_base/transformer.py", line 215, in estimate
    super(StreamingEstimationTransformer, self).estimate(X, **kwargs)
  File "/opt/miniconda3/envs/mdtools/lib/python3.9/site-packages/pyemma/coordinates/data/_base/streaming_estimator.py", line 44, in estimate
    super(StreamingEstimator, self).estimate(X, **kwargs)
  File "/opt/miniconda3/envs/mdtools/lib/python3.9/site-packages/pyemma/_base/estimator.py", line 418, in estimate
    self._model = self._estimate(X)
  File "/opt/miniconda3/envs/mdtools/lib/python3.9/site-packages/pyemma/coordinates/transform/tica.py", line 210, in _estimate
    covar.estimate(iterable, chunksize=self.chunksize, **kw)
  File "/opt/miniconda3/envs/mdtools/lib/python3.9/site-packages/pyemma/coordinates/data/_base/streaming_estimator.py", line 44, in estimate
    super(StreamingEstimator, self).estimate(X, **kwargs)
  File "/opt/miniconda3/envs/mdtools/lib/python3.9/site-packages/pyemma/_base/estimator.py", line 418, in estimate
    self._model = self._estimate(X)
  File "/opt/miniconda3/envs/mdtools/lib/python3.9/site-packages/pyemma/coordinates/estimation/covariance.py", line 240, in _estimate
    self._rc.add(X, Y, weights=weight)
  File "/opt/miniconda3/envs/mdtools/lib/python3.9/site-packages/pyemma/_ext/variational/estimators/running_moments.py", line 279, in add
    w, s_X, s_Y, C_XX, C_XY = moments_XXXY(X, Y, remove_mean=self.remove_mean, symmetrize=self.symmetrize,
  File "/opt/miniconda3/envs/mdtools/lib/python3.9/site-packages/pyemma/_ext/variational/estimators/moments.py", line 785, in moments_XXXY
    X0, mask_X, xconst, Y0, mask_Y, yconst = _sparsify_pair(X, Y, remove_mean=remove_mean, modify_data=modify_data,
  File "/opt/miniconda3/envs/mdtools/lib/python3.9/site-packages/pyemma/_ext/variational/estimators/moments.py", line 170, in _sparsify_pair
    X0, mask_X, xconst = _sparsify(X, sparse_mode=sparse_mode, sparse_tol=sparse_tol)
  File "/opt/miniconda3/envs/mdtools/lib/python3.9/site-packages/pyemma/_ext/variational/estimators/moments.py", line 149, in _sparsify
    mask = covartools.variable_cols(X, tol=sparse_tol, min_constant=min_const_col_number)  # bool vector
  File "/opt/miniconda3/envs/mdtools/lib/python3.9/site-packages/pyemma/_ext/variational/estimators/covar_c/covartools.py", line 37, in variable_cols
    cols = numpy.zeros(X.shape[1], dtype=numpy.bool, order='C')
  File "/opt/miniconda3/envs/mdtools/lib/python3.9/site-packages/numpy/__init__.py", line 284, in __getattr__
    raise AttributeError("module {!r} has no attribute "
AttributeError: module 'numpy' has no attribute 'bool'

This is with the newest numpy version
>>> numpy.__version__ '1.24.1'

The environment file is attached (env.txt) for pyemma version 2.5.12 ran on OS X v. 11.4 (Big Sur).

Minimal example

Get the trj file and structure file:

wget http://www3.mpibpc.mpg.de/groups/de_groot/compbio1/p4/md1_backbone.xtc
wget http://www3.mpibpc.mpg.de/groups/de_groot/compbio1/p4/ref.pdb
echo 0 | gmx trjconv -s ref.pdb -f md1_backbone.xtc -o md1.pdb -e 1000

Produce error:

import pyemma
import pyemma.coordinates as coor

struc = '../md1.pdb'
trj = '../md1_backbone.xtc'

# The trajectory needs to be RMSD aligned already.
feat = coor.featurizer(struc)
feat.add_all()

# Load trajectory.
inp = coor.source(trj, features=feat)

# Do the time-lagged independent component analysis (TICA)
lag = 100
tica_obj = coor.tica(inp, lag=lag, dim=2, kinetic_map=False)

Potential solution:

I edited my installation of covartools.py to replace numpy.bool with numpy.bool_ and that resolved the issue. I believe you can use the standard python bool as well. Alternatively, you could pin the numpy version in the conda install to not allow more recent numpy versions.

env.txt

[acadiaconnor]

@pitmanme I am having the same error but running within a .coordinantes.tica in my Jupyter notebook. I have tried uninstalling and reinstalling early version of numpy, with no success. Your solution worked for me! Thank you.

[pitmanme]

@acadiaconnor if you end up having other issues (I'm not using many of the functions from pyemma) you might double check which numpy loads into your Jupyter notebook.

[clonker]

Depending on what you try to do, you can also have a look at deeptime. That one is still being somewhat maintained and should be compatible to newest bumpy versions :)

[slieped]

Just came here to report the same issue.

In my case it was related to vamp, and I solve it downgrading to numpy 1.23.5 version (vs the 1.24 I was originaly using)

[clonker]

If you have a working version you could also make a PR, i'm happy to review and merge it, just don't have the time right now to do any active development (new job and all...)