Sklearn-API: better description for syntax errors

[cibic89]

First reported here.
Windows 2016 Server/Windows 10, Anaconda v2019.07, Jupyter lab v1.0.2, xgboost v0.9
Syntax errors should give better descriptions when using the sklearn API. Let me illustrate:

import pandas as pd
from sklearn.preprocessing import LabelEncoder as le
from xgboost import XGBClassifier

## Make a dummy dataframe
df1 = pd.DataFrame(
    {"x1": [0,4,2,5],
     "x2": [10,14,12,15],
     "x3": [20,24,22,25],
     "y": ["a", "b", "missing", "z"]}
)
for i in range(15):
    df1 = df1.append(df1, ignore_index = True)
print(df1.shape)
display(df1.head()) # can't remove "missing" response example as I need to prepare test set for prediction too and if I leave in NaN's label encoder will throw an error

## Make another dataframe encoding object/categorical data
df2 = df1.copy()
resp_var_le = le()
df2["y"] = resp_var_le.fit_transform(df2["y"])
print(resp_var_le.classes_)
display(df2.head())

## Prepare for xgboost
n_trees = 10
verbosity = 1
tree_method = "approx"
learning_rate = 0.3
max_depth = 6
random_state = 123
objective = 'multi:softprob'
eval_metric = "mlogloss"

## Fit a model with nothing missing
X = df2.drop(columns = ["y"])
y = df2["y"]
xgb_model = XGBClassifier(max_depth = max_depth, learning_rate = learning_rate, n_estimators = n_trees, verbosity = verbosity
                          ,objective = objective, nthread = -1, random_state = random_state, missing = resp_var_le.transform(["missing"])[0])
xgb_model.fit(X = X, y = y, eval_set = [(X, y)], eval_metric = eval_metric, early_stopping_rounds = 10)

The above code works fine but when do either change from the list below:
xgb_model.fit(X = X, y = y, eval_set = [X, y], eval_metric = eval_metric, early_stopping_rounds = 10)
or
xgb_model.fit(X = X, y = y, eval_metric = eval_metric, early_stopping_rounds = 10)
And both will give you this unintuitive error:

---------------------------------------------------------------------------
KeyError                                  Traceback (most recent call last)
~\Anaconda3\lib\site-packages\pandas\core\indexes\base.py in get_loc(self, key, method, tolerance)
   2896             try:
-> 2897                 return self._engine.get_loc(key)
   2898             except KeyError:

pandas\_libs\index.pyx in pandas._libs.index.IndexEngine.get_loc()

pandas\_libs\index.pyx in pandas._libs.index.IndexEngine.get_loc()

pandas\_libs\hashtable_class_helper.pxi in pandas._libs.hashtable.PyObjectHashTable.get_item()

pandas\_libs\hashtable_class_helper.pxi in pandas._libs.hashtable.PyObjectHashTable.get_item()

KeyError: 0

During handling of the above exception, another exception occurred:

KeyError                                  Traceback (most recent call last)
<ipython-input-1-abc953dedfab> in <module>
     46 X = df3.drop(columns = ["y"])
     47 y = df3["y"]
---> 48 xgb_model.fit(X = X, y = y, eval_set = [X, y], eval_metric = eval_metric, early_stopping_rounds = 10)

~\Anaconda3\lib\site-packages\xgboost\sklearn.py in fit(self, X, y, sample_weight, eval_set, eval_metric, early_stopping_rounds, verbose, xgb_model, sample_weight_eval_set, callbacks)
    709                         missing=self.missing, weight=sample_weight_eval_set[i],
    710                         nthread=self.n_jobs)
--> 711                 for i in range(len(eval_set))
    712             )
    713             nevals = len(evals)

~\Anaconda3\lib\site-packages\xgboost\sklearn.py in <genexpr>(.0)
    709                         missing=self.missing, weight=sample_weight_eval_set[i],
    710                         nthread=self.n_jobs)
--> 711                 for i in range(len(eval_set))
    712             )
    713             nevals = len(evals)

~\Anaconda3\lib\site-packages\pandas\core\frame.py in __getitem__(self, key)
   2978             if self.columns.nlevels > 1:
   2979                 return self._getitem_multilevel(key)
-> 2980             indexer = self.columns.get_loc(key)
   2981             if is_integer(indexer):
   2982                 indexer = [indexer]

~\Anaconda3\lib\site-packages\pandas\core\indexes\base.py in get_loc(self, key, method, tolerance)
   2897                 return self._engine.get_loc(key)
   2898             except KeyError:
-> 2899                 return self._engine.get_loc(self._maybe_cast_indexer(key))
   2900         indexer = self.get_indexer([key], method=method, tolerance=tolerance)
   2901         if indexer.ndim > 1 or indexer.size > 1:

pandas\_libs\index.pyx in pandas._libs.index.IndexEngine.get_loc()

pandas\_libs\index.pyx in pandas._libs.index.IndexEngine.get_loc()

pandas\_libs\hashtable_class_helper.pxi in pandas._libs.hashtable.PyObjectHashTable.get_item()

pandas\_libs\hashtable_class_helper.pxi in pandas._libs.hashtable.PyObjectHashTable.get_item()

KeyError: 0

[trivialfis]

We can add more checks for it like whether it's a list containing tuple, but might be too rigid. Or something like iterable... Do you have any suggestions?

[cibic89]

The easiest option is to return the native or near native error. Pretty content personally with the python API error descriptions returned.

…

________________________________ From: Jiaming Yuan <[email protected]> Sent: Thursday, September 19, 2019 5:54:40 PM To: dmlc/xgboost <[email protected]> Cc: George J <[email protected]>; Author <[email protected]> Subject: Re: [dmlc/xgboost] Sklearn-API: better description for syntax errors (#4875) We can add more checks for it like whether it's a list containing tuple, but might be too rigid. Or something like iterable... Do you have any suggestions? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#4875?email_source=notifications&email_token=AFM4L5YL7PH74O5LNSBT4ZLQKOVFBA5CNFSM4IYLVZB2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD7EEISA#issuecomment-533218376>, or mute the thread<https://github.com/notifications/unsubscribe-auth/AFM4L57SBA3GMP4UOO5OZYDQKOVFBANCNFSM4IYLVZBQ>.

[trivialfis]

@cibic89 Thanks. See if we can polish the interface more.