def _load_dataset(dataset, n_rows):
if dataset == "make_blobs":
local_X, local_y = make_blobs(n_samples=n_rows,
n_features=10,
centers=200,
cluster_std=0.8,
random_state=42)
local_X = cp.asarray(local_X)
local_y = cp.asarray(local_y)
else:
if dataset == "digits":
local_X, local_y = load_digits(return_X_y=True)
else: # dataset == "iris"
local_X, local_y = load_iris(return_X_y=True)
local_X = cp.asarray(local_X)
local_y = cp.asarray(local_y)
local_X = local_X.repeat(math.ceil(n_rows / len(local_X)), axis=0)
local_y = local_y.repeat(math.ceil(n_rows / len(local_y)), axis=0)
# Add some gaussian noise
local_X += cp.random.standard_normal(local_X.shape, dtype=cp.float32)
return local_X, local_y
def test_score(nrows, ncols, nclusters):
X, y = make_blobs(nrows,
ncols,
nclusters,
cluster_std=0.01,
random_state=10)
cuml_kmeans = cuml.KMeans(verbose=1,
init="k-means||",
n_clusters=nclusters,
random_state=10)
cuml_kmeans.fit(X)
actual_score = cuml_kmeans.score(X)
predictions = cuml_kmeans.predict(X)
centers = cp.array(cuml_kmeans.cluster_centers_.as_gpu_matrix())
expected_score = 0
for idx, label in enumerate(predictions):
x = X[idx]
y = centers[label]
dist = np.sqrt(np.sum((x - y)**2))
expected_score += dist**2
assert actual_score + SCORE_EPS \
>= (-1*expected_score) \
>= actual_score - SCORE_EPS
def test_single_linkage_sklearn_compare(nrows, ncols, nclusters, k,
connectivity):
X, y = make_blobs(int(nrows),
ncols,
nclusters,
cluster_std=1.0,
shuffle=False,
random_state=42)
cuml_agg = AgglomerativeClustering(n_clusters=nclusters,
affinity='euclidean',
linkage='single',
n_neighbors=k,
connectivity=connectivity)
try:
cuml_agg.fit(X)
except Exception:
cuml_agg.fit(X)
sk_agg = cluster.AgglomerativeClustering(n_clusters=nclusters,
affinity='euclidean',
linkage='single')
sk_agg.fit(cp.asnumpy(X))
# Cluster assignments should be exact, even though the actual
# labels may differ
assert (adjusted_rand_score(cuml_agg.labels_, sk_agg.labels_) == 1.0)
assert (cuml_agg.n_connected_components_ == sk_agg.n_connected_components_)
assert (cuml_agg.n_leaves_ == sk_agg.n_leaves_)
assert (cuml_agg.n_clusters_ == sk_agg.n_clusters_)
def test_neighborhood_predictions(nrows, ncols, n_neighbors, n_clusters,
datatype, algo):
if not has_scipy():
pytest.skip('Skipping test_neighborhood_predictions because ' +
'Scipy is missing')
X, y = make_blobs(n_samples=nrows,
centers=n_clusters,
n_features=ncols,
random_state=0)
if datatype == "dataframe":
X = cudf.DataFrame(X)
knn_cu = cuKNN(algorithm=algo)
knn_cu.fit(X)
neigh_ind = knn_cu.kneighbors(X,
n_neighbors=n_neighbors,
return_distance=False)
del knn_cu
gc.collect()
if datatype == "dataframe":
assert isinstance(neigh_ind, cudf.DataFrame)
neigh_ind = neigh_ind.to_numpy()
else:
assert isinstance(neigh_ind, cp.ndarray)
labels, probs = predict(neigh_ind, y, n_neighbors)
assert array_equal(labels, y)
def test_weighted_kmeans(nrows, ncols, nclusters, max_weight, random_state):
# Using fairly high variance between points in clusters
cluster_std = 1.0
np.random.seed(random_state)
# set weight per sample to be from 1 to max_weight
wt = np.random.randint(1, high=max_weight, size=nrows)
X, y = make_blobs(nrows,
ncols,
nclusters,
cluster_std=cluster_std,
shuffle=False,
random_state=0)
cuml_kmeans = cuml.KMeans(init="k-means++",
n_clusters=nclusters,
n_init=10,
random_state=random_state,
output_type='numpy')
cuml_kmeans.fit(X, sample_weight=wt)
cu_score = cuml_kmeans.score(X)
sk_kmeans = cluster.KMeans(random_state=random_state, n_clusters=nclusters)
sk_kmeans.fit(cp.asnumpy(X), sample_weight=wt)
sk_score = sk_kmeans.score(cp.asnumpy(X))
assert abs(cu_score - sk_score) <= cluster_std * 1.5
def test_neighborhood_predictions(nrows, ncols, n_neighbors, n_clusters,
datatype, algo):
if algo == "ivfpq":
pytest.xfail("""See Memory access error in IVFPQ :
https://github.com/rapidsai/cuml/issues/3318""")
if not has_scipy():
pytest.skip('Skipping test_neighborhood_predictions because ' +
'Scipy is missing')
X, y = make_blobs(n_samples=nrows, centers=n_clusters,
n_features=ncols, random_state=0)
if datatype == "dataframe":
X = cudf.DataFrame(X)
knn_cu = cuKNN(algorithm=algo)
knn_cu.fit(X)
neigh_ind = knn_cu.kneighbors(X, n_neighbors=n_neighbors,
return_distance=False)
del knn_cu
gc.collect()
if datatype == "dataframe":
assert isinstance(neigh_ind, cudf.DataFrame)
neigh_ind = neigh_ind.as_gpu_matrix().copy_to_host()
else:
assert isinstance(neigh_ind, cp.core.core.ndarray)
labels, probs = predict(neigh_ind, y, n_neighbors)
assert array_equal(labels, y)
def test_score(nrows, ncols, nclusters):
X, y = make_blobs(int(nrows),
ncols,
nclusters,
cluster_std=0.01,
shuffle=False,
random_state=10)
cuml_kmeans = cuml.KMeans(init="k-means||",
n_clusters=nclusters,
random_state=10,
output_type='numpy')
cuml_kmeans.fit(X)
actual_score = cuml_kmeans.score(X)
predictions = cuml_kmeans.predict(X)
centers = cuml_kmeans.cluster_centers_
expected_score = 0
for idx, label in enumerate(predictions):
x = X[idx]
y = cp.array(centers[label])
dist = cp.sqrt(cp.sum((x - y)**2))
expected_score += dist**2
assert actual_score + SCORE_EPS \
>= (-1*expected_score) \
>= actual_score - SCORE_EPS
def test_partial_fit(nrows, ncols, n_components, density, batch_size_divider,
whiten):
X, _ = make_blobs(n_samples=nrows, n_features=ncols, random_state=10)
cu_ipca = cuIPCA(n_components=n_components, whiten=whiten)
sample_size = int(nrows / batch_size_divider)
for i in range(0, nrows, sample_size):
cu_ipca.partial_fit(X[i:i + sample_size].copy())
cu_t = cu_ipca.transform(X)
cu_inv = cu_ipca.inverse_transform(cu_t)
sk_ipca = skIPCA(n_components=n_components, whiten=whiten)
X = cp.asnumpy(X)
for i in range(0, nrows, sample_size):
sk_ipca.partial_fit(X[i:i + sample_size].copy())
sk_t = sk_ipca.transform(X)
sk_inv = sk_ipca.inverse_transform(sk_t)
assert array_equal(cu_inv, sk_inv, 5e-5, with_sign=True)
def test_ivfpq_pred(nrows, ncols, n_neighbors, nlist, M, n_bits,
usePrecomputedTables):
pytest.xfail("Warning: IVFPQ might be unstable in this "
"version of cuML. This is due to a known issue "
"in the FAISS release that this cuML version "
"is linked to. (see FAISS issue #1421)")
algo_params = {
'nlist': nlist,
'nprobe': int(nlist * 0.2),
'M': M,
'n_bits': n_bits,
'usePrecomputedTables': usePrecomputedTables
}
X, y = make_blobs(n_samples=nrows,
centers=5,
n_features=ncols,
random_state=0)
knn_cu = cuKNN(algorithm="ivfpq", algo_params=algo_params)
knn_cu.fit(X)
neigh_ind = knn_cu.kneighbors(X,
n_neighbors=n_neighbors,
return_distance=False)
del knn_cu
gc.collect()
labels, probs = predict(neigh_ind, y, n_neighbors)
assert array_equal(labels, y)
def create_rand_blobs(random_state):
blobs, _ = make_blobs(n_samples=500,
n_features=20,
centers=20,
order='C',
random_state=random_state)
return blobs
def test_fit(nrows, ncols, n_components, sparse_input, density, sparse_format,
batch_size_divider, whiten):
if sparse_format == 'csc':
pytest.skip("cupyx.scipy.sparse.csc.csc_matrix does not support"
" indexing as of cupy 7.6.0")
if sparse_input:
X = cupyx.scipy.sparse.random(nrows,
ncols,
density=density,
random_state=10,
format=sparse_format)
else:
X, _ = make_blobs(n_samples=nrows, n_features=ncols, random_state=10)
cu_ipca = cuIPCA(n_components=n_components,
whiten=whiten,
batch_size=int(nrows / batch_size_divider))
cu_ipca.fit(X)
cu_t = cu_ipca.transform(X)
cu_inv = cu_ipca.inverse_transform(cu_t)
sk_ipca = skIPCA(n_components=n_components,
whiten=whiten,
batch_size=int(nrows / batch_size_divider))
if sparse_input:
X = X.get()
else:
X = cp.asnumpy(X)
sk_ipca.fit(X)
sk_t = sk_ipca.transform(X)
sk_inv = sk_ipca.inverse_transform(sk_t)
assert array_equal(cu_inv, sk_inv, 5e-5, with_sign=True)
def test_ivfsq_pred(qtype, encodeResidual, nrows, ncols, n_neighbors, nlist):
algo_params = {
'nlist': nlist,
'nprobe': nlist * 0.25,
'qtype': qtype,
'encodeResidual': encodeResidual
}
X, y = make_blobs(n_samples=nrows,
centers=5,
n_features=ncols,
random_state=0)
logger.set_level(logger.level_debug)
knn_cu = cuKNN(algorithm="ivfsq", algo_params=algo_params)
knn_cu.fit(X)
neigh_ind = knn_cu.kneighbors(X,
n_neighbors=n_neighbors,
return_distance=False)
del knn_cu
gc.collect()
labels, probs = predict(neigh_ind, y, n_neighbors)
assert array_equal(labels, y)
def test_traditional_kmeans_plus_plus_init(nrows, ncols, nclusters,
random_state):
# Using fairly high variance between points in clusters
cluster_std = 1.0
X, y = make_blobs(int(nrows),
ncols,
nclusters,
cluster_std=cluster_std,
shuffle=False,
random_state=0)
cuml_kmeans = cuml.KMeans(init="k-means++",
n_clusters=nclusters,
n_init=10,
random_state=random_state,
output_type='numpy')
cuml_kmeans.fit(X)
cu_score = cuml_kmeans.score(X)
kmeans = cluster.KMeans(random_state=random_state, n_clusters=nclusters)
kmeans.fit(cp.asnumpy(X))
sk_score = kmeans.score(cp.asnumpy(X))
assert abs(cu_score - sk_score) <= cluster_std * 1.5
def test_knn_x_none(input_type, nrows, n_feats, k, metric):
X, _ = make_blobs(n_samples=nrows,
n_features=n_feats, random_state=0)
p = 5 # Testing 5-norm of the minkowski metric only
knn_sk = skKNN(metric=metric, p=p) # Testing
knn_sk.fit(X.get())
D_sk, I_sk = knn_sk.kneighbors(X=None, n_neighbors=k)
X_orig = X
if input_type == "dataframe":
X = cudf.DataFrame(X)
knn_cu = cuKNN(metric=metric, p=p, output_type="numpy")
knn_cu.fit(X)
D_cuml, I_cuml = knn_cu.kneighbors(X=None, n_neighbors=k)
# Assert the cuml model was properly reverted
cp.testing.assert_allclose(knn_cu.X_m, X_orig,
atol=1e-5, rtol=1e-4)
# Allow a max relative diff of 10% and absolute diff of 1%
cp.testing.assert_allclose(D_cuml, D_sk, atol=5e-2,
rtol=1e-1)
assert I_cuml.all() == I_sk.all()
def test_score(nrows, ncols, nclusters, random_state):
X, y = make_blobs(int(nrows),
ncols,
nclusters,
cluster_std=1.0,
shuffle=False,
random_state=0)
cuml_kmeans = cuml.KMeans(init="k-means||",
n_clusters=nclusters,
random_state=random_state,
output_type='numpy')
cuml_kmeans.fit(X)
actual_score = cuml_kmeans.score(X)
predictions = cuml_kmeans.predict(X)
centers = cuml_kmeans.cluster_centers_
expected_score = 0.0
for idx, label in enumerate(predictions):
x = X[idx, :]
y = cp.array(centers[label, :], dtype=cp.float32)
sq_euc_dist = cp.sum(cp.square((x - y)))
expected_score += sq_euc_dist
expected_score *= -1
cp.testing.assert_allclose(actual_score,
expected_score,
atol=0.1,
rtol=1e-5)
def test_ivfpq_pred(nrows, ncols, n_neighbors, nlist, M, n_bits,
usePrecomputedTables):
algo_params = {
'nlist': nlist,
'nprobe': int(nlist * 0.2),
'M': M,
'n_bits': n_bits,
'usePrecomputedTables': usePrecomputedTables
}
X, y = make_blobs(n_samples=nrows,
centers=5,
n_features=ncols,
random_state=0)
knn_cu = cuKNN(algorithm="ivfpq", algo_params=algo_params)
knn_cu.fit(X)
neigh_ind = knn_cu.kneighbors(X,
n_neighbors=n_neighbors,
return_distance=False)
del knn_cu
gc.collect()
labels, probs = predict(neigh_ind, y, n_neighbors)
assert array_equal(labels, y)
def test_kmeans_sequential_plus_plus_init(nrows, ncols, nclusters,
random_state):
# Using fairly high variance between points in clusters
cluster_std = 1.0
X, y = make_blobs(nrows,
ncols,
nclusters,
cluster_std=cluster_std,
shuffle=False,
random_state=0)
cuml_kmeans = cuml.KMeans(verbose=0,
init="k-means++",
n_clusters=nclusters,
n_init=10,
random_state=random_state,
output_type='numpy')
cuml_kmeans.fit(X)
cu_score = cuml_kmeans.score(X)
kmeans = cluster.KMeans(random_state=random_state, n_clusters=nclusters)
kmeans.fit(X.copy_to_host())
sk_score = kmeans.score(X.copy_to_host())
assert abs(cu_score - sk_score) <= cluster_std * 1.5
def test_n_init_cluster_consistency(random_state):
cluster_std = 1.0
nrows = 100000
ncols = 100
nclusters = 8
X, y = make_blobs(nrows,
ncols,
nclusters,
cluster_std=cluster_std,
shuffle=False,
random_state=0)
cuml_kmeans = cuml.KMeans(verbose=0,
init="k-means++",
n_clusters=nclusters,
n_init=10,
random_state=random_state,
output_type='numpy')
cuml_kmeans.fit(X)
initial_clusters = cuml_kmeans.cluster_centers_
cuml_kmeans = cuml.KMeans(verbose=0,
init="k-means++",
n_clusters=nclusters,
n_init=10,
random_state=random_state,
output_type='numpy')
cuml_kmeans.fit(X)
assert array_equal(initial_clusters, cuml_kmeans.cluster_centers_)
def test_lasso_attributes():
X, y = make_blobs()
clf = cumlLasso()
clf.fit(X, y)
attrs = ["dtype", "solver_model", "coef_", "intercept_",
"solver_model", "l1_ratio", "n_cols"]
for attr in attrs:
assert hasattr(clf, attr)
def test_logistic_regression_attributes():
X, y = make_blobs()
clf = cuLog().fit(X, y, convert_dtype=True)
attrs = ["dtype", "solver_model", "coef_", "intercept_",
"l1_ratio", "n_cols", "C", "penalty",
"fit_intercept", "solver"]
for attr in attrs:
assert hasattr(clf, attr)
def test_elastic_net_attributes():
X, y = make_blobs()
clf = cumlElastic(fit_intercept=False)
clf.fit(X, y)
attrs = ["dtype", "solver_model", "coef_", "intercept_",
"l1_ratio", "n_cols", "alpha", "max_iter",
"fit_intercept"]
for attr in attrs:
assert hasattr(clf, attr)
def test_mbsgd_classifier_attributes():
X, y = make_blobs()
clf = cumlMBSGClassifier()
clf.fit(X, y)
attrs = ["dtype", "solver_model", "coef_", "intercept_",
"l1_ratio", "n_cols", "eta0", "batch_size",
"fit_intercept", "penalty"]
for attr in attrs:
assert hasattr(clf, attr)
def test_mbsgd_regressor_attributes():
X, y = make_blobs()
clf = cumlMBSGRegressor()
clf.fit(X, y)
attrs = ["dtype", "solver_model", "coef_", "intercept_",
"l1_ratio", "n_cols", "loss", "eta0", "batch_size",
"epochs"]
for attr in attrs:
assert hasattr(clf, attr)
def test_self_neighboring(datatype, metric_p, nrows):
"""Test that searches using an indexed vector itself return sensible
results for that vector
For L2-derived metrics, this specifically exercises the slow high-precision
mode used to correct for approximation errors in L2 computation during NN
searches.
"""
ncols = 1000
n_clusters = 10
n_neighbors = 3
metric, p = metric_p
if not has_scipy():
pytest.skip('Skipping test_neighborhood_predictions because ' +
'Scipy is missing')
X, y = make_blobs(n_samples=nrows,
centers=n_clusters,
n_features=ncols,
random_state=0)
if datatype == "dataframe":
X = cudf.DataFrame(X)
knn_cu = cuKNN(metric=metric, n_neighbors=n_neighbors)
knn_cu.fit(X)
neigh_dist, neigh_ind = knn_cu.kneighbors(X,
n_neighbors=n_neighbors,
return_distance=True,
two_pass_precision=True)
if datatype == 'dataframe':
assert isinstance(neigh_ind, cudf.DataFrame)
neigh_ind = neigh_ind.to_numpy()
neigh_dist = neigh_dist.to_numpy()
else:
assert isinstance(neigh_ind, cp.ndarray)
neigh_ind = neigh_ind.get()
neigh_dist = neigh_dist.get()
neigh_ind = neigh_ind[:, 0]
neigh_dist = neigh_dist[:, 0]
assert_array_equal(
neigh_ind,
np.arange(0, neigh_dist.shape[0]),
)
assert_allclose(neigh_dist,
np.zeros(neigh_dist.shape, dtype=neigh_dist.dtype),
atol=1e-4)
def get_data_consistency_test():
cluster_std = 1.0
nrows = 1000
ncols = 50
nclusters = 8
X, y = make_blobs(nrows,
ncols,
nclusters,
cluster_std=cluster_std,
shuffle=False,
random_state=0)
return X, y
def test_knn_separate_index_search(input_type, nrows, n_feats, k, metric):
X, _ = make_blobs(n_samples=nrows, n_features=n_feats, random_state=0)
X_index = X[:100]
X_search = X[101:]
p = 5 # Testing 5-norm of the minkowski metric only
knn_sk = skKNN(metric=metric, p=p) # Testing
knn_sk.fit(X_index.get())
D_sk, I_sk = knn_sk.kneighbors(X_search.get(), k)
X_orig = X_index
if input_type == "dataframe":
X_index = cudf.DataFrame(X_index)
X_search = cudf.DataFrame(X_search)
knn_cu = cuKNN(metric=metric, p=p)
knn_cu.fit(X_index)
D_cuml, I_cuml = knn_cu.kneighbors(X_search, k)
if input_type == "dataframe":
assert isinstance(D_cuml, cudf.DataFrame)
assert isinstance(I_cuml, cudf.DataFrame)
D_cuml_np = D_cuml.to_numpy()
I_cuml_np = I_cuml.to_numpy()
else:
assert isinstance(D_cuml, cp.ndarray)
assert isinstance(I_cuml, cp.ndarray)
D_cuml_np = D_cuml.get()
I_cuml_np = I_cuml.get()
with cuml.using_output_type("numpy"):
# Assert the cuml model was properly reverted
np.testing.assert_allclose(knn_cu.X_m,
X_orig.get(),
atol=1e-3,
rtol=1e-3)
if metric == 'braycurtis':
diff = D_cuml_np - D_sk
# Braycurtis has a few differences, but this is computed by FAISS.
# So long as the indices all match below, the small discrepancy
# should be okay.
assert len(diff[diff > 1e-2]) / X_search.shape[0] < 0.06
else:
np.testing.assert_allclose(D_cuml_np, D_sk, atol=1e-3, rtol=1e-3)
assert I_cuml_np.all() == I_sk.all()
def test_fuzzy_simplicial_set(n_rows, n_features, n_neighbors,
precomputed_nearest_neighbors):
n_clusters = 30
random_state = 42
metric = 'euclidean'
X, _ = make_blobs(n_samples=n_rows,
centers=n_clusters,
n_features=n_features,
random_state=random_state)
if precomputed_nearest_neighbors:
nn = NearestNeighbors(n_neighbors=n_neighbors, metric=metric)
nn.fit(X)
knn_dists, knn_indices = nn.kneighbors(X,
n_neighbors,
return_distance=True)
cu_fss_graph = cu_fuzzy_simplicial_set(X,
n_neighbors,
random_state,
metric,
knn_indices=knn_indices,
knn_dists=knn_dists)
knn_indices = knn_indices.get()
knn_dists = knn_dists.get()
ref_fss_graph = ref_fuzzy_simplicial_set(
X,
n_neighbors,
random_state,
metric,
knn_indices=knn_indices,
knn_dists=knn_dists)[0].tocoo()
else:
cu_fss_graph = cu_fuzzy_simplicial_set(X, n_neighbors, random_state,
metric)
X = X.get()
ref_fss_graph = ref_fuzzy_simplicial_set(X, n_neighbors, random_state,
metric)[0].tocoo()
cu_fss_graph = cu_fss_graph.todense()
ref_fss_graph = cp.sparse.coo_matrix(ref_fss_graph).todense()
assert correctness_sparse(ref_fss_graph,
cu_fss_graph,
atol=0.1,
rtol=0.2,
threshold=0.95)
def test_kmeans_clusters_blobs(nrows, ncols, nclusters,
random_state, cluster_std):
X, y = make_blobs(int(nrows), ncols, nclusters,
cluster_std=cluster_std,
shuffle=False,
random_state=random_state,)
cuml_kmeans = cuml.KMeans(init="k-means||",
n_clusters=nclusters,
random_state=random_state,
output_type='numpy')
preds = cuml_kmeans.fit_predict(X)
assert adjusted_rand_score(cp.asnumpy(preds), cp.asnumpy(y)) >= 0.99
def test_return_dists():
n_samples = 50
n_feats = 50
k = 5
X, y = make_blobs(n_samples=n_samples, n_features=n_feats, random_state=0)
knn_cu = cuKNN()
knn_cu.fit(X)
ret = knn_cu.kneighbors(X, k, return_distance=False)
assert not isinstance(ret, tuple)
assert ret.shape == (n_samples, k)
ret = knn_cu.kneighbors(X, k, return_distance=True)
assert isinstance(ret, tuple)
assert len(ret) == 2
def test_ann_distances_metrics(algo, metric):
X, y = make_blobs(n_samples=500, centers=2, n_features=128, random_state=0)
cu_knn = cuKNN(algorithm=algo, metric=metric)
cu_knn.fit(X)
cu_dist, cu_ind = cu_knn.kneighbors(X,
n_neighbors=10,
return_distance=True)
del cu_knn
gc.collect()
X = X.get()
sk_knn = skKNN(metric=metric)
sk_knn.fit(X)
sk_dist, sk_ind = sk_knn.kneighbors(X,
n_neighbors=10,
return_distance=True)
return array_equal(sk_dist, cu_dist)