DataCanvasIO / HyperTS

A Full-Pipeline Automated Time Series (AutoTS) Analysis Toolkit.
https://hyperts.readthedocs.io
Apache License 2.0
260 stars 27 forks source link

Support scipy >=1.8.0 #70

Open DManowitz opened 1 year ago

DManowitz commented 1 year ago

Right now, this package only supports scipy<1.8.0. However, newer versions are available.

zhangxjohn commented 1 year ago

Right now, this package only supports scipy<1.8.0. However, newer versions are available.

Thanks, this issue will be solved in the next release.

DManowitz commented 1 year ago

I see that the requirements.txt has changed to reflect this, and in conda-forge, the dependencies no longer has a restriction on scipy version. However, in conda-forge, it still lists scipy < 1.8.0 as a constraint. I also see that version 0.2.0 has added a constraint on arviz <= 0.12.1 on conda-forge. Why has that been added?

oaksharks commented 1 year ago

Hi @DManowitz , The constrains specified in run_constrained only affect the building runtime environment, not as dependencies of the package. Do you have any problems about installing hyperts ? If it is, you can further describe it and list your python environment.

References:

DManowitz commented 1 year ago

When I view the hyperts 0.2.0 package information on conda-forge, I see the following:

hyperts 0.2.0 pyhd8ed1ab_3
--------------------------
file name   : hyperts-0.2.0-pyhd8ed1ab_3.conda
name        : hyperts
version     : 0.2.0
build       : pyhd8ed1ab_3
build number: 3
size        : 780 KB
license     : Apache-2.0
subdir      : noarch
url         : https://conda.anaconda.org/conda-forge/noarch/hyperts-0.2.0-pyhd8ed1ab_3.conda
md5         : 99658aa2ba533cf003658204b52ab5fb
timestamp   : 2022-11-24 04:06:52 UTC
constraints :
  - tensorflow >=2.0
  - scipy <1.8.0
  - arviz <=0.12.1
dependencies:
  - hypernets >=0.2.5.1,<0.2.6
  - numpy >=1.16.5,<1.22
  - pandas >=0.25.3
  - prophet >=1.0.1
  - python >=3.6
  - scikit-learn >=0.22.1
  - scipy
  - sktime >=0.8.1
  - statsmodels >=0.12.1

As specified here and here , the constraints in the run_constrained section "allows defining restrictions on packages at runtime." Since scipy is already a dependency, also having the constraint of scipy <1.8.0 effectively is the same as having a dependency of scipy <1.8.0.