AutoGluon | 用三行代码战胜 90% 的模型

生信菜鸟团

发布于 2021-05-24 10:29:49

8.2K1

发布于 2021-05-24 10:29:49

近年来，机器学习在各个领域都取得了重大突破，在生命科学、医学领域应用的也越来越多。但想要真正建立一个模型仍费时费力，要花相当一段时间去学习（可参考我之前写的相关笔记）。而且即使是高水平的人工智能专家，在大数据智能分析机器学习建模时，也主要依靠人工经验，建模过程费时费力，缺少有效方法。

为了解决这一突出问题，自动机器学习（AutoML）应运而生，即是将机器学习应用于现实问题的端到端过程自动化的过程。AutoML 使机器学习真正意义上成为可能，即使对于在该领域没有专业知识的人也是如此。

本文介绍的 AutoGluon 是去年亚马逊开源在 GitHub 上的 AutoML 解决方案，它以三年前亚马逊联合微软推出的 Gluon 深度学习库为基础，通过自动调整限制条件内的选择来实现决策的自动化，利用现有资源找到最优的模型。

❝GitHub 地址：https://github.com/awslabs/autogluon ❞

AutoGluon 兼具易用和扩展性，并专注于涵盖图像、文本或表格数据的深度学习和实际应用。AutoGluon 适用于机器学习初学者和专家，能够使他们：

通过几行代码即可快速地为数据构建深度学习原型方案；
利用自动超参数微调、模型选择/架构搜索和数据处理；
无需专家知识即可自动使用深度学习 SOTA 方法；
轻松地提升现有定制模型和数据管道，或者根据用例自定义 AutoGluon。

现在 AutoGluon 已经支持了以下一些应用：

表格预测：基于数据表中一些列的值预测其他列的值；
图像分类：识别图像中的主要对象；
对象检测：借助图像中的边界框检测多个对象；
文本分类：基于文本内容做出预测。

怎样使用 AutoGluon 呢？官方文档中提供了各种应用的示例代码，下面我节选了最常用的 Tablular Prediction。

❝官方文档：https://auto.gluon.ai/stable/index.html ❞

安装

Linux

python3 -m pip install -U pip
python3 -m pip install -U setuptools wheel
python3 -m pip install -U "mxnet<2.0.0"
python3 -m pip install autogluon

python3 -m pip install -U pip
python3 -m pip install -U setuptools wheel

# Here we assume CUDA 10.1 is installed.  You should change the number
# according to your own CUDA version (e.g. mxnet_cu100 for CUDA 10.0).
python3 -m pip install -U "mxnet_cu101<2.0.0"
python3 -m pip install autogluon

Mac

❝

需先安装 XCode，Homebrew，LibOMP。若已经安装了 Homebrew，则可通过以下命令安装 LibOMP：

brew install libomp

❞

python3 -m pip install -U pip
python3 -m pip install -U setuptools wheel
python3 -m pip install -U "mxnet<2.0.0"
python3 -m pip install autogluon

三行代码构建一个绝佳的模型

# 载入包
from autogluon.tabular import TabularDataset, TabularPredictor
# 载入训练数据
train_data = TabularDataset('https://autogluon.s3.amazonaws.com/datasets/Inc/train.csv')
# 建模
predictor = TabularPredictor(label='class').fit(train_data, time_limit=120)  # Fit models for 120s
# 载入测试数据
test_data = TabularDataset('https://autogluon.s3.amazonaws.com/datasets/Inc/test.csv')
# 查看模型性能排名
leaderboard = predictor.leaderboard(test_data)

基于表格数据预测

通过一个简单的 fit() 函数，AutoGluon 即可生成高度精确的模型，本例演示了如何使用 AutoGluon 生成一个分类模型来预测一个人的收入是否超过 50,000 美元。

首先，导入 AutoGluon 的 TabularPredictor 和 TabularDataset 类:

from autogluon.tabular import TabularDataset, TabularPredictor

将训练数据从 CSV 文件加载到 AutoGluon Dataset 对象中。这个对象实质上等同于 Pandas DataFrame。

train_data = TabularDataset('https://autogluon.s3.amazonaws.com/datasets/Inc/train.csv')
subsample_size = 500  # subsample subset of data for faster demo, try setting this to much larger values
train_data = train_data.sample(n=subsample_size, random_state=0)
train_data.head()

train_data 中的每一行都对应一个样本，每一列包含各种特征，我们将使用这些特征来预测收入。首先指定 'class' 列为标签：

label = 'class'
print("Summary of class variable: \n", train_data[label].describe())

Summary of class variable:
 count        500
unique         2
top        <=50K
freq         365
Name: class, dtype: object

接下来使用 AutoGluon 自动训练模型：

# 指定储存模型的文件夹
save_path = 'agModels-predictClass'
predictor = TabularPredictor(label=label, path=save_path).fit(train_data)

Beginning AutoGluon training ...
AutoGluon will save models to "agModels-predictClass/"
AutoGluon Version:  0.2.0b20210429
Train Data Rows:    500
Train Data Columns: 14
Preprocessing data ...
AutoGluon infers your prediction problem is: 'binary' (because only two unique label-values observed).
    2 unique label values:  [' >50K', ' <=50K']
    If 'binary' is not the correct problem_type, please manually specify the problem_type argument in fit() (You may specify problem_type as one of: ['binary', 'multiclass', 'regression'])
Selected class <--> label mapping:  class 1 =  >50K, class 0 =  <=50K
    Note: For your binary classification, AutoGluon arbitrarily selected which label-value represents positive ( >50K) vs negative ( <=50K) class.
    To explicitly set the positive_class, either rename classes to 1 and 0, or specify positive_class in Predictor init.
Using Feature Generators to preprocess the data ...
Fitting AutoMLPipelineFeatureGenerator...
    Available Memory:                    26825.14 MB
    Train Data (Original)  Memory Usage: 0.29 MB (0.0% of available memory)
    Inferring data type of each feature based on column values. Set feature_metadata_in to manually specify special dtypes of the features.
    Stage 1 Generators:
            Fitting AsTypeFeatureGenerator...
    Stage 2 Generators:
            Fitting FillNaFeatureGenerator...
    Stage 3 Generators:
            Fitting IdentityFeatureGenerator...
            Fitting CategoryFeatureGenerator...
                    Fitting CategoryMemoryMinimizeFeatureGenerator...
    Stage 4 Generators:
            Fitting DropUniqueFeatureGenerator...
    Types of features in original data (raw dtype, special dtypes):
            ('int', [])    : 6 | ['age', 'fnlwgt', 'education-num', 'capital-gain', 'capital-loss', ...]
            ('object', []) : 8 | ['workclass', 'education', 'marital-status', 'occupation', 'relationship', ...]
    Types of features in processed data (raw dtype, special dtypes):
            ('category', []) : 8 | ['workclass', 'education', 'marital-status', 'occupation', 'relationship', ...]
            ('int', [])      : 6 | ['age', 'fnlwgt', 'education-num', 'capital-gain', 'capital-loss', ...]
    0.1s = Fit runtime
    14 features in original data used to generate 14 features in processed data.
    Train Data (Processed) Memory Usage: 0.03 MB (0.0% of available memory)
Data preprocessing and feature engineering runtime = 0.07s ...
AutoGluon will gauge predictive performance using evaluation metric: 'accuracy'
    To change this, specify the eval_metric argument of fit()
Automatically generating train/validation split with holdout_frac=0.2, Train Rows: 400, Val Rows: 100
Fitting model: KNeighborsUnif ...
    0.73     = Validation accuracy score
    0.0s     = Training runtime
    0.1s     = Validation runtime
Fitting model: KNeighborsDist ...
    0.65     = Validation accuracy score
    0.0s     = Training runtime
    0.1s     = Validation runtime
Fitting model: LightGBMXT ...
/var/lib/jenkins/workspace/workspace/autogluon-tutorial-tabular-v3/venv/lib/python3.7/site-packages/fsspec/__init__.py:47: DeprecationWarning: SelectableGroups dict interface is deprecated. Use select.
  for spec in entry_points.get("fsspec.specs", []):
    0.83     = Validation accuracy score
    0.77s    = Training runtime
    0.01s    = Validation runtime
Fitting model: LightGBM ...
    0.85     = Validation accuracy score
    0.16s    = Training runtime
    0.01s    = Validation runtime
Fitting model: RandomForestGini ...
    0.84     = Validation accuracy score
    0.51s    = Training runtime
    0.11s    = Validation runtime
Fitting model: RandomForestEntr ...
    0.83     = Validation accuracy score
    0.5s     = Training runtime
    0.11s    = Validation runtime
Fitting model: CatBoost ...
    0.84     = Validation accuracy score
    0.39s    = Training runtime
    0.01s    = Validation runtime
Fitting model: ExtraTreesGini ...
    0.82     = Validation accuracy score
    0.5s     = Training runtime
    0.11s    = Validation runtime
Fitting model: ExtraTreesEntr ...
    0.82     = Validation accuracy score
    0.5s     = Training runtime
    0.11s    = Validation runtime
Fitting model: NeuralNetFastAI ...
/var/lib/jenkins/workspace/workspace/autogluon-tutorial-tabular-v3/venv/lib/python3.7/site-packages/torch/cuda/__init__.py:52: UserWarning: CUDA initialization: The NVIDIA driver on your system is too old (found version 10010). Please update your GPU driver by downloading and installing a new version from the URL: http://www.nvidia.com/Download/index.aspx Alternatively, go to: https://pytorch.org to install a PyTorch version that has been compiled with your version of the CUDA driver. (Triggered internally at  /pytorch/c10/cuda/CUDAFunctions.cpp:100.)
  return torch._C._cuda_getDeviceCount() > 0
No improvement since epoch 7: early stopping
    0.83     = Validation accuracy score
    1.15s    = Training runtime
    0.03s    = Validation runtime
Fitting model: XGBoost ...
    0.85     = Validation accuracy score
    0.17s    = Training runtime
    0.01s    = Validation runtime
Fitting model: NeuralNetMXNet ...
    0.84     = Validation accuracy score
    5.57s    = Training runtime
    0.02s    = Validation runtime
Fitting model: LightGBMLarge ...
    0.83     = Validation accuracy score
    0.37s    = Training runtime
    0.01s    = Validation runtime
Fitting model: WeightedEnsemble_L2 ...
    0.85     = Validation accuracy score
    0.35s    = Training runtime
    0.0s     = Validation runtime
AutoGluon training complete, total runtime = 12.68s ...
TabularPredictor saved. To load, use: predictor = TabularPredictor.load("agModels-predictClass/")

下一步，加载独立测试数据集：

test_data = TabularDataset('https://autogluon.s3.amazonaws.com/datasets/Inc/test.csv')
y_test = test_data[label]  # 提取标签列
test_data_nolab = test_data.drop(columns=[label])  # 删除标签列
test_data_nolab.head()

Loaded data from: https://autogluon.s3.amazonaws.com/datasets/Inc/test.csv | Columns = 15 / 15 | Rows = 9769 -> 9769

用训练好的模型进行预测，评估模型性能：

predictor = TabularPredictor.load(save_path)  # 非必要步骤，载入之前训练好的模型

y_pred = predictor.predict(test_data_nolab)
print("Predictions:  \n", y_pred)
perf = predictor.evaluate_predictions(y_true=y_test, y_pred=y_pred, auxiliary_metrics=True)

Evaluation: accuracy on test data: 0.8397993653393387
Evaluations on test data:
{
    "accuracy": 0.8397993653393387,
    "balanced_accuracy": 0.7437076677780596,
    "mcc": 0.5295565206264157,
    "f1": 0.6242496998799519,
    "precision": 0.7038440714672441,
    "recall": 0.5608283002588438
}

Predictions:
 0        <=50K
1        <=50K
2         >50K
3        <=50K
4        <=50K
         ...
9764     <=50K
9765     <=50K
9766     <=50K
9767     <=50K
9768     <=50K
Name: class, Length: 9769, dtype: object

我们也可以输出所有模型在测试集中的表现：

predictor.leaderboard(test_data, silent=True)

总之，如果想在自己的数据集上使用 AutoGluon，训练一个牛逼的模型实际只需两行代码：

from autogluon.tabular import TabularPredictor
predictor = TabularPredictor(label=<variable-name>).fit(train_data=<file-name>)

❝上面的例子仅仅使用了 fit() 的默认参数，在下一节中将演示如何通过调整 fit() 参数来最大化预测性能。 ❞

最大限度地提高预测性能

如果你正在对 AutoGluon-Tabular 进行基准测试或希望最大限度地提高模型准确性，其实不应使用 fit() 的默认参数。为了获得 AutoGluon 的最佳预测精度，你通常应该使用下面的几行命令：

time_limit = 60  # 这里设置 60 秒仅用于快速演示代码，在实际中应该设置为你愿意等待的最长时间（以秒为单位）
metric = 'roc_auc'  # 指定模型性能指标，这里为 AUC
predictor = TabularPredictor(label, eval_metric=metric).fit(train_data, time_limit=time_limit, presets='best_quality')
predictor.leaderboard(test_data, silent=True)

No path specified. Models will be saved in: "AutogluonModels/ag-20210429_010727/"
Presets specified: ['best_quality']
Beginning AutoGluon training ... Time limit = 60s
AutoGluon will save models to "AutogluonModels/ag-20210429_010727/"
AutoGluon Version:  0.2.0b20210429
Train Data Rows:    500
Train Data Columns: 14
Preprocessing data ...
AutoGluon infers your prediction problem is: 'binary' (because only two unique label-values observed).
    2 unique label values:  [' >50K', ' <=50K']
    If 'binary' is not the correct problem_type, please manually specify the problem_type argument in fit() (You may specify problem_type as one of: ['binary', 'multiclass', 'regression'])
Selected class <--> label mapping:  class 1 =  >50K, class 0 =  <=50K
    Note: For your binary classification, AutoGluon arbitrarily selected which label-value represents positive ( >50K) vs negative ( <=50K) class.
    To explicitly set the positive_class, either rename classes to 1 and 0, or specify positive_class in Predictor init.
Using Feature Generators to preprocess the data ...
Fitting AutoMLPipelineFeatureGenerator...
    Available Memory:                    26582.18 MB
    Train Data (Original)  Memory Usage: 0.29 MB (0.0% of available memory)
    Inferring data type of each feature based on column values. Set feature_metadata_in to manually specify special dtypes of the features.
    Stage 1 Generators:
            Fitting AsTypeFeatureGenerator...
    Stage 2 Generators:
            Fitting FillNaFeatureGenerator...
    Stage 3 Generators:
            Fitting IdentityFeatureGenerator...
            Fitting CategoryFeatureGenerator...
                    Fitting CategoryMemoryMinimizeFeatureGenerator...
    Stage 4 Generators:
            Fitting DropUniqueFeatureGenerator...
    Types of features in original data (raw dtype, special dtypes):
            ('int', [])    : 6 | ['age', 'fnlwgt', 'education-num', 'capital-gain', 'capital-loss', ...]
            ('object', []) : 8 | ['workclass', 'education', 'marital-status', 'occupation', 'relationship', ...]
    Types of features in processed data (raw dtype, special dtypes):
            ('category', []) : 8 | ['workclass', 'education', 'marital-status', 'occupation', 'relationship', ...]
            ('int', [])      : 6 | ['age', 'fnlwgt', 'education-num', 'capital-gain', 'capital-loss', ...]
    0.1s = Fit runtime
    14 features in original data used to generate 14 features in processed data.
    Train Data (Processed) Memory Usage: 0.03 MB (0.0% of available memory)
Data preprocessing and feature engineering runtime = 0.07s ...
AutoGluon will gauge predictive performance using evaluation metric: 'roc_auc'
    This metric expects predicted probabilities rather than predicted class labels, so you'll need to use predict_proba() instead of predict()
    To change this, specify the eval_metric argument of fit()
Fitting model: KNeighborsUnif_BAG_L1 ... Training model for up to 59.93s of the 59.92s of remaining time.
    0.5196   = Validation roc_auc score
    0.0s     = Training runtime
    0.1s     = Validation runtime
Fitting model: KNeighborsDist_BAG_L1 ... Training model for up to 59.82s of the 59.82s of remaining time.
    0.537    = Validation roc_auc score
    0.0s     = Training runtime
    0.1s     = Validation runtime
Fitting model: LightGBMXT_BAG_L1 ... Training model for up to 59.71s of the 59.71s of remaining time.
    0.8814   = Validation roc_auc score
    0.9s     = Training runtime
    0.05s    = Validation runtime
Fitting model: LightGBM_BAG_L1 ... Training model for up to 58.74s of the 58.73s of remaining time.
    0.867    = Validation roc_auc score
    0.91s    = Training runtime
    0.05s    = Validation runtime
Fitting model: RandomForestGini_BAG_L1 ... Training model for up to 57.75s of the 57.75s of remaining time.
    0.8847   = Validation roc_auc score
    0.5s     = Training runtime
    0.09s    = Validation runtime
Fitting model: RandomForestEntr_BAG_L1 ... Training model for up to 57.15s of the 57.14s of remaining time.
    0.8863   = Validation roc_auc score
    0.5s     = Training runtime
    0.09s    = Validation runtime
Fitting model: CatBoost_BAG_L1 ... Training model for up to 56.54s of the 56.54s of remaining time.
    0.8875   = Validation roc_auc score
    2.93s    = Training runtime
    0.04s    = Validation runtime
Fitting model: ExtraTreesGini_BAG_L1 ... Training model for up to 53.55s of the 53.55s of remaining time.
    0.8929   = Validation roc_auc score
    0.6s     = Training runtime
    0.09s    = Validation runtime
Fitting model: ExtraTreesEntr_BAG_L1 ... Training model for up to 52.83s of the 52.83s of remaining time.
    0.8939   = Validation roc_auc score
    0.5s     = Training runtime
    0.09s    = Validation runtime
Fitting model: NeuralNetFastAI_BAG_L1 ... Training model for up to 52.22s of the 52.22s of remaining time.
    0.8653   = Validation roc_auc score
    4.69s    = Training runtime
    0.13s    = Validation runtime
Fitting model: XGBoost_BAG_L1 ... Training model for up to 47.34s of the 47.33s of remaining time.
    0.8666   = Validation roc_auc score
    1.84s    = Training runtime
    0.03s    = Validation runtime
Fitting model: NeuralNetMXNet_BAG_L1 ... Training model for up to 45.41s of the 45.41s of remaining time.
    Ran out of time, stopping training early. (Stopping on epoch 95)
    0.8331   = Validation roc_auc score
    31.16s   = Training runtime
    0.18s    = Validation runtime
Fitting model: LightGBMLarge_BAG_L1 ... Training model for up to 14.02s of the 14.02s of remaining time.
    0.8417   = Validation roc_auc score
    1.7s     = Training runtime
    0.05s    = Validation runtime
Completed 1/20 k-fold bagging repeats ...
Fitting model: WeightedEnsemble_L2 ... Training model for up to 59.93s of the 12.24s of remaining time.
    0.9003   = Validation roc_auc score
    1.26s    = Training runtime
    0.0s     = Validation runtime
AutoGluon training complete, total runtime = 49.04s ...
TabularPredictor saved. To load, use: predictor = TabularPredictor.load("AutogluonModels/ag-20210429_010727/")

这行命令正是用了下面的策略以最大限度地提高模型准确性：

指定参数 presets='best_quality'，这使 AutoGluon 可以基于 stacking/bagging 自动构建功能强大的集成模型，如果有足够的训练时间这将大大提高预测性能。这个参数的默认值为 'medium_quality_faster_train'，它产生的模型准确度较低，但可以更快地建立一个模型，查看分类性能。总之，使用 presets 参数，我们可以根据用途灵活地建立模型。比如，如果你只是希望快速构建一个基本模型，并不非常注重预测性能，则可以考虑使用：presets=['good_quality_faster_inference_only_refit', 'optimize_for_deployment']。
eval_metric 参数用于设置模型评估指标，例如：'f1'（二分类），'roc_auc'（二分类），'log_loss'（分类），'mean_absolute_error'（回归），'median_absolute_error'（回归）。除此之外，我们自定义指标函数，具体参见文件夹中的示例：autogluon/core/metrics/
将所有数据都包含在 train_data 中，不额外分配tuning_data（AutoGluon 会更智能地拆分数据以满足其需求）。
不指定 hyperparameter_tune_kwargs 参数（与直觉相反，超参数调整并非是在有限时间内训练模型的最佳方法，集成模型一般会更好）。如果你只是想部署单个模型而不是集成模型，则建议仅使用hyperparameter_tune_kwargs。
不指定 hyperparameters 参数（AutoGluon 将自适应选择要使用的模型和超参数）。
将 time_limit 设置的尽可能长。

回归（预测连续性变量）

为了演示 fit() 也可以自动处理回归任务，我们现在尝试根据其他特性预测同一数据集中的年龄变量：

age_column = 'age'
print("Summary of age variable: \n", train_data[age_column].describe())

Summary of age variable:
 count    500.00000
mean      39.65200
std       13.52393
min       17.00000
25%       29.00000
50%       38.00000
75%       49.00000
max       85.00000
Name: age, dtype: float64

我们再次调用 fit()，这次也加了一个时间限制：

predictor_age = TabularPredictor(label=age_column, path="agModels-predictAge").fit(train_data, time_limit=60)
performance = predictor_age.evaluate(test_data)

Beginning AutoGluon training ... Time limit = 60s
AutoGluon will save models to "agModels-predictAge/"
AutoGluon Version:  0.2.0b20210429
Train Data Rows:    500
Train Data Columns: 14
Preprocessing data ...
AutoGluon infers your prediction problem is: 'regression' (because dtype of label-column == int and many unique label-values observed).
    Label info (max, min, mean, stddev): (85, 17, 39.652, 13.52393)
    If 'regression' is not the correct problem_type, please manually specify the problem_type argument in fit() (You may specify problem_type as one of: ['binary', 'multiclass', 'regression'])
Using Feature Generators to preprocess the data ...
Fitting AutoMLPipelineFeatureGenerator...
    Available Memory:                    26519.77 MB
    Train Data (Original)  Memory Usage: 0.32 MB (0.0% of available memory)
    Inferring data type of each feature based on column values. Set feature_metadata_in to manually specify special dtypes of the features.
    Stage 1 Generators:
            Fitting AsTypeFeatureGenerator...
    Stage 2 Generators:
            Fitting FillNaFeatureGenerator...
    Stage 3 Generators:
            Fitting IdentityFeatureGenerator...
            Fitting CategoryFeatureGenerator...
                    Fitting CategoryMemoryMinimizeFeatureGenerator...
    Stage 4 Generators:
            Fitting DropUniqueFeatureGenerator...
    Types of features in original data (raw dtype, special dtypes):
            ('int', [])    : 5 | ['fnlwgt', 'education-num', 'capital-gain', 'capital-loss', 'hours-per-week']
            ('object', []) : 9 | ['workclass', 'education', 'marital-status', 'occupation', 'relationship', ...]
    Types of features in processed data (raw dtype, special dtypes):
            ('category', []) : 9 | ['workclass', 'education', 'marital-status', 'occupation', 'relationship', ...]
            ('int', [])      : 5 | ['fnlwgt', 'education-num', 'capital-gain', 'capital-loss', 'hours-per-week']
    0.1s = Fit runtime
    14 features in original data used to generate 14 features in processed data.
    Train Data (Processed) Memory Usage: 0.03 MB (0.0% of available memory)
Data preprocessing and feature engineering runtime = 0.07s ...
AutoGluon will gauge predictive performance using evaluation metric: 'root_mean_squared_error'
    To change this, specify the eval_metric argument of fit()
Automatically generating train/validation split with holdout_frac=0.2, Train Rows: 400, Val Rows: 100
Fitting model: KNeighborsUnif ... Training model for up to 59.93s of the 59.93s of remaining time.
    -15.6869         = Validation root_mean_squared_error score
    0.0s     = Training runtime
    0.1s     = Validation runtime
Fitting model: KNeighborsDist ... Training model for up to 59.82s of the 59.82s of remaining time.
    -15.1801         = Validation root_mean_squared_error score
    0.0s     = Training runtime
    0.1s     = Validation runtime
Fitting model: LightGBMXT ... Training model for up to 59.72s of the 59.72s of remaining time.
    -11.8147         = Validation root_mean_squared_error score
    0.21s    = Training runtime
    0.01s    = Validation runtime
Fitting model: LightGBM ... Training model for up to 59.49s of the 59.48s of remaining time.
    -11.9295         = Validation root_mean_squared_error score
    0.19s    = Training runtime
    0.01s    = Validation runtime
Fitting model: RandomForestMSE ... Training model for up to 59.28s of the 59.28s of remaining time.
    -11.6028         = Validation root_mean_squared_error score
    0.5s     = Training runtime
    0.11s    = Validation runtime
Fitting model: CatBoost ... Training model for up to 58.66s of the 58.66s of remaining time.
    -11.7448         = Validation root_mean_squared_error score
    0.34s    = Training runtime
    0.01s    = Validation runtime
Fitting model: ExtraTreesMSE ... Training model for up to 58.31s of the 58.31s of remaining time.
    -11.4808         = Validation root_mean_squared_error score
    0.5s     = Training runtime
    0.11s    = Validation runtime
Fitting model: NeuralNetFastAI ... Training model for up to 57.69s of the 57.68s of remaining time.
    -40.9923         = Validation root_mean_squared_error score
    0.97s    = Training runtime
    0.03s    = Validation runtime
Fitting model: XGBoost ... Training model for up to 56.68s of the 56.68s of remaining time.
    -12.1743         = Validation root_mean_squared_error score
    0.53s    = Training runtime
    0.01s    = Validation runtime
Fitting model: NeuralNetMXNet ... Training model for up to 56.12s of the 56.12s of remaining time.
    -12.8081         = Validation root_mean_squared_error score
    6.81s    = Training runtime
    0.03s    = Validation runtime
Fitting model: LightGBMLarge ... Training model for up to 49.27s of the 49.27s of remaining time.
    -12.1676         = Validation root_mean_squared_error score
    0.51s    = Training runtime
    0.01s    = Validation runtime
Fitting model: WeightedEnsemble_L2 ... Training model for up to 59.93s of the 48.14s of remaining time.
    -11.3341         = Validation root_mean_squared_error score
    0.42s    = Training runtime
    0.0s     = Validation runtime
AutoGluon training complete, total runtime = 12.3s ...
TabularPredictor saved. To load, use: predictor = TabularPredictor.load("agModels-predictAge/")
Evaluation: root_mean_squared_error on test data: -10.459079691277005
    Note: Scores are always higher_is_better. This metric score can be multiplied by -1 to get the metric value.
Evaluations on test data:
{
    "root_mean_squared_error": -10.459079691277005,
    "mean_squared_error": -109.39234798848308,
    "mean_absolute_error": -8.194572974525316,
    "r2": 0.4152766647850106,
    "pearsonr": 0.6454293362505439,
    "median_absolute_error": -6.8994293212890625
}

这里我们不需要告诉 AutoGluon 这是一个回归问题，它会从数据中自动进行推断，并报告适当的性能指标（默认情况下为 RMSE）。

查看每个模型的性能：