ValueError: feature_names mismatch:
最近测试平台算子,发现xgb算子出现bug:feature_names mismatch
ValueError: feature_names mismatch: ['a1', 'a2', 'a3', 'a4'] ['f0', 'f1', 'f2', 'f3'] expected a1, a3, a2, a4 in input data training data did not have the following fields: f2, f1, f3, f0
分析一番后,终于找到原因,这是xgb算子本身的bug,在训练时候数据的列顺序被打乱了,所以在predict的时候报出顺序不一致的错误
解决办法:加入validate_features=False参数
model.predict(X,validate_features=False)