Description
Description
While trying to roll my own estimator as part of kernelmethods, I ran into a failed estimator check check_non_transformer_estimators_n_iter.
This function is not supposed to run the check (as my estimator is derived from SVR) according to code below:
# These models are dependent on external solvers like
# libsvm and accessing the iter parameter is non-trivial.
not_run_check_n_iter = ['Ridge', 'SVR', 'NuSVR', 'NuSVC',
'RidgeClassifier', 'SVC', 'RandomizedLasso',
'LogisticRegressionCV', 'LinearSVC',
'LogisticRegression']
# Tested in test_transformer_n_iter
not_run_check_n_iter += CROSS_DECOMPOSITION
if name in not_run_check_n_iter:
returnHowever, given my estimator name (OptimalKernelSVR) is not actually in the list, it goes ahead and runs the check which is irrelevant to my estimator. To prevent this, I suggest making these checks less specific to sklearn's native estimators and more amenable to external libraries. Some possible solutions are:
- change
if name in not_run_check_n_iter:to
for no_test_name in not_run_check_n_iter:
if name.lower() in no_test_name.lower():
return - Allow passing some flags to
check_estimatorskip a specific test in here
Steps/Code to Reproduce
-
Clone and install kernelmethods in dev mode
-
Run this code
from kernelmethods.algorithms import OptimalKernelSVR
from sklearn.utils.estimator_checks import check_estimator
OKSVR = OptimalKernelSVR(k_bucket='light')
check_estimator(OKSVR)Versions
System:
python: 3.7.2 (default, Dec 29 2018, 00:00:04) [Clang 4.0.1 (tags/RELEASE_401/final)]
executable: /Users/Reddy/anaconda3/envs/py36/bin/python
machine: Darwin-18.6.0-x86_64-i386-64bit
BLAS:
macros: SCIPY_MKL_H=None, HAVE_CBLAS=None
lib_dirs: /Users/Reddy/anaconda3/envs/py36/lib
cblas_libs: mkl_rt, pthread
Python deps:
pip: 19.1.1
setuptools: 41.0.1
sklearn: 0.21.2
numpy: 1.15.4
scipy: 1.1.0
Cython: None
pandas: 0.24.2