def test_cast_double_to_float():
DT = dt.Frame(F=[0.78, math.nan, math.inf, -math.inf, 2.35e78, -6.5e11])
assert DT.stypes == (dt.float64,)
RES = DT[:, dt.float32(f.F)]
assert RES.stypes == (dt.float32,)
assert RES.to_list()[0] == [0.7799999713897705, None, math.inf, -math.inf,
math.inf, -6.50000007168e+11]
def test_cast_to_float32(src):
dt0 = dt.Frame(src)
dt1 = dt0[:, [dt.float32(f[i]) for i in range(dt0.ncols)]]
dt1.internal.check()
assert dt1.stypes == (dt.float32, ) * dt0.ncols
pyans = [float(x) if x is not None else None for x in src]
assert list_equals(dt1.topython()[0], pyans)
def test_ftrl_reuse_pickled_empty_model():
ft_pickled = pickle.dumps(Ftrl())
ft_unpickled = pickle.loads(ft_pickled)
ft_unpickled.nbins = 10
df_train = dt.Frame({"id": range(ft_unpickled.nbins)})
df_target = dt.Frame([True] * ft_unpickled.nbins)
ft_unpickled.fit(df_train, df_target)
model = [[-0.5] * ft_unpickled.nbins, [0.25] * ft_unpickled.nbins]
fi = dt.Frame([["id"], [0.0]])[:, [f[0], dt.float32(f[1])]]
fi.names = ["feature_name", "feature_importance"]
assert ft_unpickled.model.to_list() == model
assert_equals(ft_unpickled.feature_importances, fi)
def test_cast_view_simple():
df0 = dt.Frame({"A": [1, 2, 3]})
df1 = df0[::-1, :][:, dt.float32(f.A)]
frame_integrity_check(df1)
assert df1.stypes == (dt.float32, )
assert df1.to_list() == [[3.0, 2.0, 1.0]]
def test_cast_view():
df0 = dt.Frame({"A": [1, 2, 3]})
df1 = df0[::-1, :][:, dt.float32(f.A)]
df1.internal.check()
assert df1.stypes == (dt.float32, )
assert df1.topython() == [[3.0, 2.0, 1.0]]
def transform(self, X: dt.Frame):
if dtype_global() == np.float32:
return X[:, [dt.float32(dt.log(dt.f[i])) for i in range(X.ncols)]]
else:
return X[:, [dt.float64(dt.log(dt.f[i])) for i in range(X.ncols)]]
