def test_ols(nrows, ncols, n_parts, fit_intercept, normalize, datatype,
delayed, client):
def imp():
import cuml.comm.serialize # NOQA
client.run(imp)
from cuml.dask.linear_model import LinearRegression as cumlOLS_dask
n_info = 5
nrows = int(nrows)
ncols = int(ncols)
X, y = make_regression_dataset(datatype, nrows, ncols, n_info)
X_df, y_df = _prep_training_data(client, X, y, n_parts)
lr = cumlOLS_dask(fit_intercept=fit_intercept, normalize=normalize)
lr.fit(X_df, y_df)
ret = lr.predict(X_df, delayed=delayed)
error_cuml = mean_squared_error(y, ret.compute().to_pandas().values)
assert (error_cuml < 1e-6)
def test_ols(cluster):
client = Client(cluster)
try:
import dask_cudf
import cudf
import numpy as np
from cuml.dask.linear_model import LinearRegression as cumlOLS_dask
nrows = 2**8
ncols = 399
X, y = load_data(nrows, ncols)
X_cudf = cudf.DataFrame.from_pandas(X)
y_cudf = np.array(y.as_matrix())
y_cudf = y_cudf[:, 0]
y_cudf = cudf.Series(y_cudf)
workers = client.has_what().keys()
X_df = dask_cudf.from_cudf(X_cudf, npartitions=len(workers)).persist()
y_df = dask_cudf.from_cudf(y_cudf, npartitions=len(workers)).persist()
lr = cumlOLS_dask()
lr.fit(X_df, y_df)
ret = lr.predict(X_df)
error_cuml = mean_squared_error(y, ret.compute().to_array())
assert(error_cuml < 1e-6)
finally:
client.close()
cluster.close()
def test_ols(nrows,
ncols,
n_parts,
fit_intercept,
normalize,
datatype,
client=None):
if client is None:
cluster = LocalCUDACluster()
client = Client(cluster)
try:
from cuml.dask.linear_model import LinearRegression as cumlOLS_dask
n_info = 5
nrows = np.int(nrows)
ncols = np.int(ncols)
X, y = make_regression_dataset(datatype, nrows, ncols, n_info)
X_df, y_df = _prep_training_data(client, X, y, n_parts)
lr = cumlOLS_dask(fit_intercept=fit_intercept, normalize=normalize)
if n_parts > 2:
lr.fit(X_df, y_df, force_colocality=True)
else:
lr.fit(X_df, y_df)
ret = lr.predict(X_df)
error_cuml = mean_squared_error(y, ret.compute().to_pandas().values)
assert (error_cuml < 1e-6)
finally:
client.close()
cluster.close()