[hotfix] Fix the issue mentioned in #352

[nabenabe0928]

Types of changes

• Bug fix (non-breaking change which fixes an issue)

Checklist:

• My code follows the code style of this project.
• My change requires a change to the documentation.
• I have updated the documentation accordingly.

• Have you checked to ensure there aren't other open Pull Requests for the same update/change?
• Have you added an explanation of what your changes do and why you'd like us to include them?
• Have you written new tests for your core changes, as applicable?
• Have you successfully ran tests with your changes locally?

Description

This is the hotfix for the issue#352.
Since the inference of the task type in sklearn does not properly handle the continuous, I added a feature to filter continuous (i.e. regression task) from multiclass.

Motivation and Context

see issue#352..

How has this been tested?

It is not tested by proper pytest, but I checked by the following locally.
Note that data.csv is available in issue#352.

from autoPyTorch.api.tabular_regression import TabularRegressionTask
import pandas as pd
from sklearn.model_selection import train_test_split


input = [
    "P1",
    "P5p1",
    "P6p2",
    "P11p4",
    "P14p9",
    "P15p1",
    "P15p3",
    "P16p2",
    "P18p2",
    "P27p4",
    "H2p2",
    "H8p2",
    "H10p1",
    "H13p1",
    "H18pA",
    "H40p4",
]
output = ["price"]

if __name__ == '__main__':
    # read dataframe
    df = pd.read_csv("data.csv")

    # split data
    X = df[input]
    y = df[output]
    X_train, X_test, y_train, y_test = train_test_split(
        X, y, train_size=0.8, random_state=1
    )

    # train model
    api = TabularRegressionTask(n_jobs=7)
    api.search(
        X_train=X_train,
        y_train=y_train,
        X_test=X_test.copy(),
        y_test=y_test.copy(),
        optimize_metric="r2",
        total_walltime_limit=120,
    )

Note that the original code yields the following error by the same code:

Traceback (most recent call last):
  File "bug_test.py", line 39, in <module>
    api.search(
  File "/home/Auto-PyTorch/autoPyTorch/api/tabular_regression.py", line 300, in search
    return self._search(
  File "/home/Auto-PyTorch/autoPyTorch/api/base_task.py", line 876, in _search
    raise ValueError("Incompatible dataset entered for current task,"
ValueError: Incompatible dataset entered for current task,expected dataset to have task type :tabular_regression got :tabular_classification

[codecov]

Codecov Report

Merging #354 (8545155) into development (8f9e9f6) will increase coverage by 0.01%.
The diff coverage is 80.00%.

@@               Coverage Diff               @@
##           development     #354      +/-   ##
===============================================
+ Coverage        82.62%   82.64%   +0.01%     
===============================================
  Files              154      154              
  Lines             9055     9076      +21     
  Branches          1592     1598       +6     
===============================================
+ Hits              7482     7501      +19     
- Misses            1106     1107       +1     
- Partials           467      468       +1     

Impacted Files	Coverage Δ
autoPyTorch/datasets/base_dataset.py	80.00% <80.00%> (-0.80%)	⬇️
autoPyTorch/api/base_task.py	83.60% <0.00%> (ø)
autoPyTorch/utils/results_visualizer.py	99.05% <0.00%> (+0.05%)	⬆️
...peline/components/training/trainer/base_trainer.py	96.82% <0.00%> (+1.05%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 8f9e9f6...8545155. Read the comment docs.

[nabenabe0928]

This change is the core of this PR, but for now since the original author does not response to the issue anymore, we close this PR.

Note that the issue happened because the sklearn task inference for multiclass is incorrect, i.e. if the target is only int, it will mess up the target with multiclass even if it is really continuou.