xgboost bug (handling nested array)

bug.py

# Version: 0.6a2
from xgboost import XGBRegressor

model = XGBRegressor(seed = 1)

# XOR pattern + [0, 0, 0]
train_data = [
    [1, 0] + [0] * 3,
    [0, 1] + [0] * 3,
    [1, 1] + [0] * 3,
    [0, 0] + [0] * 3,
]

train_y = [1, 1, 0, 0]

model.fit(train_data, train_y)

# XOR Pattern + [1, 1, 1]
test_data = [
    [0, 0] + [1] * 3
]

model.predict(test_data)
# ValueError: feature_names mismatch: ['f0', 'f1'] ['f0', 'f1', 'f2', 'f3', 'f4']