def test_new_img_like_side_effect():
img1 = Nifti1Image(np.ones((2, 2, 2, 2)), affine=np.eye(4))
hash1 = joblib.hash(img1)
new_img_like(img1, np.ones((2, 2, 2, 2)), img1.affine.copy(),
copy_header=True)
hash2 = joblib.hash(img1)
assert_equal(hash1, hash2)
def test_joblib_cache():
if not LooseVersion(nibabel.__version__) > LooseVersion('1.1.0'):
# Old nibabel do not pickle
raise SkipTest
from sklearn.externals.joblib import hash, Memory
mask = np.zeros((40, 40, 40))
mask[20, 20, 20] = 1
mask_img = Nifti1Image(mask, np.eye(4))
with testing.write_tmp_imgs(mask_img, create_files=True)\
as filename:
masker = NiftiMasker(mask_img=filename)
masker.fit()
mask_hash = hash(masker.mask_img_)
masker.mask_img_.get_data()
assert_true(mask_hash == hash(masker.mask_img_))
# Test a tricky issue with memmapped joblib.memory that makes
# imgs return by inverse_transform impossible to save
cachedir = mkdtemp()
try:
masker.memory = Memory(cachedir=cachedir, mmap_mode='r',
verbose=0)
X = masker.transform(mask_img)
# inverse_transform a first time, so that the result is cached
out_img = masker.inverse_transform(X)
out_img = masker.inverse_transform(X)
out_img.to_filename(os.path.join(cachedir, 'test.nii'))
finally:
shutil.rmtree(cachedir, ignore_errors=True)
def hash_X_y(X, y, n_samples=10, n_features=5):
"""Compute hash of the input arrays.
Parameters
----------
X : ndarray, shape (n_samples, n_features)
The ``X`` array.
y : ndarray, shape (n_samples)
The ``y`` array.
n_samples : int, optional
The number of samples to use to compute the hash. Default is 100.
n_features : int, optional
The number of features to use to compute the hash. Default is 10.
Returns
-------
X_hash: str
Hash identifier of the ``X`` matrix.
y_hash: str
Hash identifier of the ``y`` matrix.
"""
row_idx = slice(None, None, max(1, X.shape[0] // n_samples))
col_idx = slice(None, None, max(1, X.shape[1] // n_features))
return joblib.hash(X[row_idx, col_idx]), joblib.hash(y[row_idx])
def test_new_img_like_side_effect():
img1 = Nifti1Image(np.ones((2, 2, 2, 2)), affine=np.eye(4))
hash1 = joblib.hash(img1)
new_img_like(img1, np.ones((2, 2, 2, 2)), img1.get_affine().copy(),
copy_header=True)
hash2 = joblib.hash(img1)
assert_equal(hash1, hash2)
def check_estimators_overwrite_params(name, Estimator):
X, y = make_blobs(random_state=0, n_samples=9)
y = multioutput_estimator_convert_y_2d(name, y)
# some want non-negative input
X -= X.min()
with warnings.catch_warnings(record=True):
# catch deprecation warnings
estimator = Estimator()
set_fast_parameters(estimator)
set_random_state(estimator)
# Make a physical copy of the orginal estimator parameters before fitting.
params = estimator.get_params()
original_params = deepcopy(params)
# Fit the model
estimator.fit(X, y)
# Compare the state of the model parameters with the original parameters
new_params = estimator.get_params()
for param_name, original_value in original_params.items():
new_value = new_params[param_name]
# We should never change or mutate the internal state of input
# parameters by default. To check this we use the joblib.hash function
# that introspects recursively any subobjects to compute a checksum.
# The only exception to this rule of immutable constructor parameters
# is possible RandomState instance but in this check we explicitly
# fixed the random_state params recursively to be integer seeds.
assert_equal(hash(new_value), hash(original_value),
"Estimator %s should not change or mutate "
" the parameter %s from %s to %s during fit."
% (name, param_name, original_value, new_value))
def test_joblib_cache():
if not LooseVersion(nibabel.__version__) > LooseVersion('1.1.0'):
# Old nibabel do not pickle
raise SkipTest
from sklearn.externals.joblib import hash, Memory
mask = np.zeros((40, 40, 40))
mask[20, 20, 20] = 1
mask_img = Nifti1Image(mask, np.eye(4))
with testing.write_tmp_imgs(mask_img, create_files=True)\
as filename:
masker = NiftiMasker(mask_img=filename)
masker.fit()
mask_hash = hash(masker.mask_img_)
masker.mask_img_.get_data()
assert_true(mask_hash == hash(masker.mask_img_))
# Test a tricky issue with memmapped joblib.memory that makes
# imgs return by inverse_transform impossible to save
cachedir = mkdtemp()
try:
masker.memory = Memory(cachedir=cachedir, mmap_mode='r', verbose=0)
X = masker.transform(mask_img)
# inverse_transform a first time, so that the result is cached
out_img = masker.inverse_transform(X)
out_img = masker.inverse_transform(X)
out_img.to_filename(os.path.join(cachedir, 'test.nii'))
finally:
shutil.rmtree(cachedir, ignore_errors=True)
def check_estimators_overwrite_params(name, Estimator):
X, y = make_blobs(random_state=0, n_samples=9)
y = multioutput_estimator_convert_y_2d(name, y)
# some want non-negative input
X -= X.min()
with warnings.catch_warnings(record=True):
# catch deprecation warnings
estimator = Estimator()
set_fast_parameters(estimator)
set_random_state(estimator)
# Make a physical copy of the orginal estimator parameters before fitting.
params = estimator.get_params()
original_params = deepcopy(params)
# Fit the model
estimator.fit(X, y)
# Compare the state of the model parameters with the original parameters
new_params = estimator.get_params()
for param_name, original_value in original_params.items():
new_value = new_params[param_name]
# We should never change or mutate the internal state of input
# parameters by default. To check this we use the joblib.hash function
# that introspects recursively any subobjects to compute a checksum.
# The only exception to this rule of immutable constructor parameters
# is possible RandomState instance but in this check we explicitly
# fixed the random_state params recursively to be integer seeds.
assert_equal(
hash(new_value), hash(original_value),
"Estimator %s should not change or mutate "
" the parameter %s from %s to %s during fit." %
(name, param_name, original_value, new_value))
def hash_X_y(X, y, n_samples=10, n_features=5):
"""Compute hash of the input arrays.
Parameters
----------
X : ndarray, shape (n_samples, n_features)
The ``X`` array.
y : ndarray, shape (n_samples)
The ``y`` array.
n_samples : int, optional
The number of samples to use to compute the hash. Default is 100.
n_features : int, optional
The number of features to use to compute the hash. Default is 10.
Returns
-------
X_hash: str
Hash identifier of the ``X`` matrix.
y_hash: str
Hash identifier of the ``y`` matrix.
"""
row_idx = slice(None, None, max(1, X.shape[0] // n_samples))
col_idx = slice(None, None, max(1, X.shape[1] // n_features))
return joblib.hash(X[row_idx, col_idx]), joblib.hash(y[row_idx])
def test_check_estimator_clones():
# check that check_estimator doesn't modify the estimator it receives
from sklearn.datasets import load_iris
iris = load_iris()
for Estimator in [GaussianMixture, LinearRegression,
RandomForestClassifier, NMF, SGDClassifier,
MiniBatchKMeans]:
with ignore_warnings(category=FutureWarning):
# when 'est = SGDClassifier()'
est = Estimator()
set_checking_parameters(est)
set_random_state(est)
# without fitting
old_hash = joblib.hash(est)
check_estimator(est)
assert_equal(old_hash, joblib.hash(est))
with ignore_warnings(category=FutureWarning):
# when 'est = SGDClassifier()'
est = Estimator()
set_checking_parameters(est)
set_random_state(est)
# with fitting
est.fit(iris.data + 10, iris.target)
old_hash = joblib.hash(est)
check_estimator(est)
assert_equal(old_hash, joblib.hash(est))
def _check_integrity_atlas(atlas):
# check that the folder is existing
atlas_directory = os.path.abspath(os.path.join('.', 'data', 'fmri', atlas))
if os.path.isdir(atlas_directory):
# compute the hash of the current data present on the disk
filenames_atlas_current = np.array(glob.glob(
os.path.join(atlas_directory, '*', '*', '*')),
dtype=object)
filenames_atlas_current.sort()
current_hash = joblib.hash(filenames_atlas_current)
# compute the expected hash from the data set which we provide
filenames_atlas_expected = pd.read_csv(os.path.abspath(
os.path.join('.', 'data', 'fmri_filename.csv')),
index_col=0)[atlas].values
for idx in range(filenames_atlas_expected.size):
filenames_atlas_expected[idx] = os.path.abspath(
filenames_atlas_expected[idx])
filenames_atlas_expected.sort()
expected_hash = joblib.hash(filenames_atlas_expected)
if current_hash == expected_hash:
return
shutil.rmtree(atlas_directory)
_download_fmri_data(atlas)
def test_hash_X_y():
rng = check_random_state(0)
X = rng.randn(2000, 20)
y = np.array([0] * 500 + [1] * 1500)
assert hash_X_y(X, y, 10,
10) == (joblib.hash(X[::200, ::2]), joblib.hash(y[::200]))
X = rng.randn(5, 2)
y = np.array([0] * 2 + [1] * 3)
# all data will be used in this case
assert hash_X_y(X, y) == (joblib.hash(X), joblib.hash(y))
def test_hash_X_y():
rng = check_random_state(0)
X = rng.randn(2000, 20)
y = np.array([0] * 500 + [1] * 1500)
assert hash_X_y(X, y, 10, 10) == (joblib.hash(X[::200, ::2]),
joblib.hash(y[::200]))
X = rng.randn(5, 2)
y = np.array([0] * 2 + [1] * 3)
# all data will be used in this case
assert hash_X_y(X, y) == (joblib.hash(X), joblib.hash(y))
def test_hash_X_y_pandas():
pd = pytest.importorskip("pandas")
rng = check_random_state(0)
X = pd.DataFrame(rng.randn(2000, 20))
y = pd.Series([0] * 500 + [1] * 1500)
assert hash_X_y(X, y, 10, 10) == (joblib.hash(X.iloc[::200, ::2]),
joblib.hash(y.iloc[::200]))
X = pd.DataFrame(rng.randn(5, 2))
y = pd.Series([0] * 2 + [1] * 3)
# all data will be used in this case
assert hash_X_y(X, y) == (joblib.hash(X), joblib.hash(y))
def test_joblib_cache():
from sklearn.externals.joblib import hash
# Dummy mask
mask = np.zeros((40, 40, 40))
mask[20, 20, 20] = 1
mask_img = Nifti1Image(mask, np.eye(4))
with write_tmp_imgs(mask_img, create_files=True) as filename:
masker = MultiNiftiMasker(mask_img=filename)
masker.fit()
mask_hash = hash(masker.mask_img_)
masker.mask_img_.get_data()
assert_true(mask_hash == hash(masker.mask_img_))
# enables to delete "filename" on windows
del masker
def __init__(self, tmp_dir, array):
h = hash(array)
hashs[h] = array
output_filename = os.path.join(tmp_dir, h)
with lockfile.LockFile(output_filename):
if not os.path.exists(output_filename):
dump(array, output_filename)
self.filename = output_filename
def test_joblib_cache():
if not LooseVersion(nibabel.__version__) > LooseVersion('1.1.0'):
# Old nibabel do not pickle
raise SkipTest
from sklearn.externals.joblib import hash
# Dummy mask
mask = np.zeros((40, 40, 40))
mask[20, 20, 20] = 1
mask_img = Nifti1Image(mask, np.eye(4))
with write_tmp_imgs(mask_img, create_files=True)\
as filename:
masker = MultiNiftiMasker(mask_img=filename)
masker.fit()
mask_hash = hash(masker.mask_img_)
masker.mask_img_.get_data()
assert_true(mask_hash == hash(masker.mask_img_))
def test_joblib_cache():
if not LooseVersion(nibabel.__version__) > LooseVersion('1.1.0'):
# Old nibabel do not pickle
raise SkipTest
from sklearn.externals.joblib import hash
# Dummy mask
mask = np.zeros((40, 40, 40))
mask[20, 20, 20] = 1
mask_img = Nifti1Image(mask, np.eye(4))
with testing.write_tmp_imgs(mask_img, create_files=True)\
as filename:
masker = MultiNiftiMasker(mask_img=filename)
masker.fit()
mask_hash = hash(masker.mask_img_)
masker.mask_img_.get_data()
assert_true(mask_hash == hash(masker.mask_img_))
def evaluate_one(model_class, parameters, cv_split):
split_idx, (X_train, X_val, y_train, y_val) = cv_split
model = model_class(**parameters).fit(X_train, y_train)
train_score = model.score(X_train, y_train)
validation_score = model.score(X_val, y_val)
results = {
'train_score': train_score,
'val_score': validation_score,
'parameters': parameters,
'parameters_hash': hash(parameters),
}
return results
def hash_X_y(X, y, n_samples=1000):
"""Compute hash of the input arrays.
Parameters
----------
X : ndarray, shape (n_samples, n_features)
The ``X`` array.
y : ndarray, shape (n_samples)
Returns
-------
X_hash: str
Hash identifier of the ``X`` matrix.
y_hash: str
Hash identifier of the ``y`` matrix.
"""
rng = np.random.RandomState(0)
raw_idx = rng.randint(X.shape[0], size=n_samples)
col_idx = rng.randint(X.shape[1], size=n_samples)
return joblib.hash(X[raw_idx, col_idx]), joblib.hash(y[raw_idx])
def evaluate_one(model_class, parameters, cv_split):
split_idx, (X_train, X_val, y_train, y_val) = cv_split
model = model_class(**parameters).fit(X_train, y_train)
train_score = model.score(X_train, y_train)
validation_score = model.score(X_val, y_val)
results = {
'train_score': train_score,
'val_score': validation_score,
'parameters': parameters,
'parameters_hash': hash(parameters),
}
return results
def test_copy_img_side_effect():
img1 = Nifti1Image(np.ones((2, 2, 2, 2)), affine=np.eye(4))
hash1 = joblib.hash(img1)
niimg.copy_img(img1)
hash2 = joblib.hash(img1)
assert_equal(hash1, hash2)
joblib.dump((X_vali, y_vali, qid_vali), VALI_DATA)
if not os.path.exists(GRID_JOBS_FOLDER):
os.makedirs(GRID_JOBS_FOLDER)
params = {
'max_features': [10, 20, 50, 100],
'max_depth': [2, 3, 4, 5],
'subsample': [0.5, 0.8, 1.0],
'loss': ['ls', 'huber', 'quantile'],
'learning_rate': [0.05, 0.1, 0.5],
}
for i, param in enumerate(ParameterGrid(params)):
params_description = json.dumps(param)
job_id = joblib.hash(params_description)
job_folder = GRID_JOBS_FOLDER + '/' + job_id
if not os.path.exists(job_folder):
os.makedirs(job_folder)
with open(job_folder + '/parameters.json', 'wb') as f:
f.write(params_description.encode('utf-8'))
data_filenames = {'train': TRAIN_SAMPLE_DATA, 'validation': VALI_DATA}
with open(job_folder + '/data.json', 'wb') as f:
f.write(json.dumps(data_filenames).encode('utf-8'))
cmd = 'qsub -V -cwd letor_gridpoint.py {}'.format(job_folder)
os.system(cmd)
# if i > 100:
# break
def __hash__(self):
if self._hashvalue is None:
self._hashvalue = int(joblib.hash(self._hashkey), 16)
return self._hashvalue
def test_copy_img_side_effect():
img1 = Nifti1Image(np.ones((2, 2, 2, 2)), affine=np.eye(4))
hash1 = joblib.hash(img1)
niimg.copy_img(img1)
hash2 = joblib.hash(img1)
assert_equal(hash1, hash2)
if not os.path.exists(GRID_JOBS_FOLDER):
os.makedirs(GRID_JOBS_FOLDER)
params = {
'max_features': [10, 20, 50, 100],
'max_depth': [2, 3, 4, 5],
'subsample': [0.5, 0.8, 1.0],
'loss': ['ls', 'huber', 'quantile'],
'learning_rate': [0.05, 0.1, 0.5],
}
for i, param in enumerate(ParameterGrid(params)):
params_description = json.dumps(param)
job_id = joblib.hash(params_description)
job_folder = GRID_JOBS_FOLDER + '/' + job_id
if not os.path.exists(job_folder):
os.makedirs(job_folder)
with open(job_folder + '/parameters.json', 'wb') as f:
f.write(params_description.encode('utf-8'))
data_filenames = {'train': TRAIN_SAMPLE_DATA, 'validation': VALI_DATA}
with open(job_folder + '/data.json', 'wb') as f:
f.write(json.dumps(data_filenames).encode('utf-8'))
cmd = 'qsub -V -cwd letor_gridpoint.py {}'.format(job_folder)
os.system(cmd)
# if i > 100:
# break
