def test_contextualbanditlabel_multiple_label():
df = pd.DataFrame(
{"a1": [1], "c1": [-0.5], "p1": [0.1], "a2": [2], "c2": [-1.5], "p2": [0.6], "x": [1]}
)
conv = DFtoVW(
df=df,
label=[ContextualbanditLabel("a1", "c1", "p1"), ContextualbanditLabel("a2", "c2", "p2")],
features=Feature("x"),
)
first_line = conv.convert_df()[0]
assert first_line == "1:-0.5:0.1 2:-1.5:0.6 | x:1"
def test_listlabel_mixed_label_error():
with pytest.raises(TypeError) as type_error:
df = pd.DataFrame({"a": [1], "c": [-0.5], "p": [0.1], "b": [2], "x": [1]})
conv = DFtoVW(
df=df,
label=[ContextualbanditLabel("a", "c", "p"), MultiLabel("b")],
features=Feature("x"),
)
expected = "The list passed in 'label' has mixed label types."
assert expected == str(type_error.value)
def test_contextualbanditlabel_non_positive_action():
df = pd.DataFrame({"a": [0], "c": [-0.5], "p": [0.5], "x": [1]})
with pytest.raises(ValueError) as value_error:
DFtoVW(
df=df,
label=ContextualbanditLabel("a", "c", "p"),
features=Feature("x"),
)
expected = "In argument 'action' of 'ContextualbanditLabel', column 'a' must be >= 1."
assert expected == str(value_error.value)
def test_contextualbanditlabel_non_float_proba_error():
df = pd.DataFrame({"a": [1], "c": [-0.5], "p": [1], "x": [1]})
with pytest.raises(TypeError) as value_error:
DFtoVW(
df=df,
label=ContextualbanditLabel("a", "c", "p"),
features=Feature("x"),
)
expected = "In argument 'probability' of 'ContextualbanditLabel', column 'p' should be either of the following type(s): 'float'."
assert expected == str(value_error.value)
def test_contextualbanditlabel_negative_proba_error():
df = pd.DataFrame({"a": [1], "c": [-0.5], "p": [-0.1], "x": [1]})
with pytest.raises(ValueError) as value_error:
DFtoVW(
df=df,
label=ContextualbanditLabel("a", "c", "p"),
features=Feature("x"),
)
expected = (
"In argument 'probability' of 'ContextualbanditLabel', column 'p' must be >= 0 and <= 1."
)
assert expected == str(value_error.value)
def test_contextualbanditlabel_one_label():
df = pd.DataFrame({"a": [1], "c": [-0.5], "p": [0.1], "x": [1]})
conv = DFtoVW(df=df, label=ContextualbanditLabel("a", "c", "p"), features=Feature("x"))
first_line = conv.convert_df()[0]
assert first_line == "1:-0.5:0.1 | x:1"