10BC0 FIX Preserve float32 in MiniBatchDictionaryLearning by jeremiedbb · Pull Request #22428 · scikit-learn/scikit-learn · GitHub
[go: up one dir, main page]
Skip to content

FIX Preserve float32 in MiniBatchDictionaryLearning #22428

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
ogrisel merged 2 commits into from
Feb 10, 2022
Merged

FIX Preserve float32 in MiniBatchDictionaryLearning #22428

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
d456d2c
preserve dtype for inner_stats
jeremiedbb Feb 9, 2022
f80342f
cln
jeremiedbb Feb 9, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
19 changes: 9 additions & 10 deletions sklearn/decomposition/_dict_learning.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -893,7 +893,8 @@ def dict_learning_online(
dictionary = dictionary[:n_components, :]
else:
dictionary = np.r_[
dictionary, np.zeros((n_components - r, dictionary.shape[1]))
dictionary,
np.zeros((n_components - r, dictionary.shape[1]), dtype=dictionary.dtype),
]

if verbose == 1:
Expand All @@ -905,25 +906,23 @@ def dict_learning_online(
else:
X_train = X

# Fortran-order dict better suited for the sparse coding which is the
# bottleneck of this algorithm.
dictionary = check_array(
dictionary, order="F", dtype=[np.float64, np.float32], copy=False
)
dictionary = np.require(dictionary, requirements="W")

X_train = check_array(
X_train, order="C", dtype=[np.float64, np.float32], copy=False
)

# Fortran-order dict better suited for the sparse coding which is the
# bottleneck of this algorithm.
dictionary = check_array(dictionary, order="F", dtype=X_train.dtype, copy=False)
dictionary = np.require(dictionary, requirements="W")

batches = gen_batches(n_samples, batch_size)
batches = itertools.cycle(batches)

# The covariance of the dictionary
if inner_stats is None:
A = np.zeros((n_components, n_components))
A = np.zeros((n_components, n_components), dtype=X_train.dtype)
# The data approximation
B = np.zeros((n_features, n_components))
B = np.zeros((n_features, n_components), dtype=X_train.dtype)
else:
A = inner_stats[0].copy()
B = inner_stats[1].copy()
Expand Down
4 changes: 4 additions & 0 deletions sklearn/decomposition/tests/test_dict_learning.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -740,6 +740,10 @@ def test_dictionary_learning_dtype_match(
assert dict_learner.components_.dtype == expected_type
assert dict_learner.transform(X.astype(data_type)).dtype == expected_type

if dictionary_learning_transformer is MiniBatchDictionaryLearning:
assert dict_learner.inner_stats_[0].dtype == expected_type
assert dict_learner.inner_stats_[1].dtype == expected_type


@pytest.mark.parametrize("method", ("lars", "cd"))
@pytest.mark.parametrize(
Expand Down
0