Predicting Columns in a Table - In Depth

Tip: If you are new to AutoGluon, review Predicting Columns in a Table - Quick Start to learn the basics of the AutoGluon API.

This tutorial describes how you can exert greater control when using AutoGluon’s fit() or predict(). Recall that to maximize predictive performance, you should always first try fit() with all default arguments except eval_metric and presets, before you experiment with other arguments covered in this in-depth tutorial like hyperparameter_tune_kwargs, hyperparameters, num_stack_levels, num_bag_folds, num_bag_sets, etc.

Using the same census data table as in the Predicting Columns in a Table - Quick Start tutorial, we’ll now predict the occupation of an individual - a multiclass classification problem. Start by importing AutoGluon’s TabularPredictor and TabularDataset, and loading the data.

from autogluon.tabular import TabularDataset, TabularPredictor

import numpy as np

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)
print(train_data.head())

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

new_data = TabularDataset('https://autogluon.s3.amazonaws.com/datasets/Inc/test.csv')
test_data = new_data[5000:].copy()  # this should be separate data in your applications
y_test = test_data[label]
test_data_nolabel = test_data.drop(columns=[label])  # delete label column
val_data = new_data[:5000].copy()

metric = 'accuracy' # we specify eval-metric just for demo (unnecessary as it's the default)
       age workclass  fnlwgt      education  education-num  6118    51   Private   39264   Some-college             10
23204   58   Private   51662           10th              6
29590   40   Private  326310   Some-college             10
18116   37   Private  222450        HS-grad              9
33964   62   Private  109190      Bachelors             13

            marital-status        occupation    relationship    race      sex  6118    Married-civ-spouse   Exec-managerial            Wife   White   Female
23204   Married-civ-spouse     Other-service            Wife   White   Female
29590   Married-civ-spouse      Craft-repair         Husband   White     Male
18116        Never-married             Sales   Not-in-family   White     Male
33964   Married-civ-spouse   Exec-managerial         Husband   White     Male

       capital-gain  capital-loss  hours-per-week  native-country   class
6118              0             0              40   United-States    >50K
23204             0             0               8   United-States   <=50K
29590             0             0              44   United-States   <=50K
18116             0          2339              40     El-Salvador   <=50K
33964         15024             0              40   United-States    >50K
Summary of occupation column:
 count                  500
unique                  15
top        Exec-managerial
freq                    77
Name: occupation, dtype: object

Specifying hyperparameters and tuning them

We first demonstrate hyperparameter-tuning and how you can provide your own validation dataset that AutoGluon internally relies on to: tune hyperparameters, early-stop iterative training, and construct model ensembles. One reason you may specify validation data is when future test data will stem from a different distribution than training data (and your specified validation data is more representative of the future data that will likely be encountered).

If you don’t have a strong reason to provide your own validation dataset, we recommend you omit the tuning_data argument. This lets AutoGluon automatically select validation data from your provided training set (it uses smart strategies such as stratified sampling). For greater control, you can specify the holdout_frac argument to tell AutoGluon what fraction of the provided training data to hold out for validation.

Caution: Since AutoGluon tunes internal knobs based on this validation data, performance estimates reported on this data may be over-optimistic. For unbiased performance estimates, you should always call predict() on a separate dataset (that was never passed to fit()), as we did in the previous Quick-Start tutorial. We also emphasize that most options specified in this tutorial are chosen to minimize runtime for the purposes of demonstration and you should select more reasonable values in order to obtain high-quality models.

fit() trains neural networks and various types of tree ensembles by default. You can specify various hyperparameter values for each type of model. For each hyperparameter, you can either specify a single fixed value, or a search space of values to consider during hyperparameter optimization. Hyperparameters which you do not specify are left at default settings chosen automatically by AutoGluon, which may be fixed values or search spaces.

import autogluon.core as ag

nn_options = {  # specifies non-default hyperparameter values for neural network models
    'num_epochs': 10,  # number of training epochs (controls training time of NN models)
    'learning_rate': ag.space.Real(1e-4, 1e-2, default=5e-4, log=True),  # learning rate used in training (real-valued hyperparameter searched on log-scale)
    'activation': ag.space.Categorical('relu', 'softrelu', 'tanh'),  # activation function used in NN (categorical hyperparameter, default = first entry)
    'layers': ag.space.Categorical([100], [1000], [200, 100], [300, 200, 100]),  # each choice for categorical hyperparameter 'layers' corresponds to list of sizes for each NN layer to use
    'dropout_prob': ag.space.Real(0.0, 0.5, default=0.1),  # dropout probability (real-valued hyperparameter)
}

gbm_options = {  # specifies non-default hyperparameter values for lightGBM gradient boosted trees
    'num_boost_round': 100,  # number of boosting rounds (controls training time of GBM models)
    'num_leaves': ag.space.Int(lower=26, upper=66, default=36),  # number of leaves in trees (integer hyperparameter)
}

hyperparameters = {  # hyperparameters of each model type
                   'GBM': gbm_options,
                   'NN': nn_options,  # NOTE: comment this line out if you get errors on Mac OSX
                  }  # When these keys are missing from hyperparameters dict, no models of that type are trained

time_limit = 2*60  # train various models for ~2 min
num_trials = 5  # try at most 5 different hyperparameter configurations for each type of model
search_strategy = 'auto'  # to tune hyperparameters using Bayesian optimization routine with a local scheduler

hyperparameter_tune_kwargs = {  # HPO is not performed unless hyperparameter_tune_kwargs is specified
    'num_trials': num_trials,
    'scheduler' : 'local',
    'searcher': search_strategy,
}

predictor = TabularPredictor(label=label, eval_metric=metric).fit(
    train_data, tuning_data=val_data, time_limit=time_limit,
    hyperparameters=hyperparameters, hyperparameter_tune_kwargs=hyperparameter_tune_kwargs,
)
No path specified. Models will be saved in: "AutogluonModels/ag-20210617_032450/"
Warning: hyperparameter tuning is currently experimental and may cause the process to hang.
Beginning AutoGluon training ... Time limit = 120s
AutoGluon will save models to "AutogluonModels/ag-20210617_032450/"
AutoGluon Version:  0.2.1b20210617
Train Data Rows:    500
Train Data Columns: 14
Tuning Data Rows:    5000
Tuning Data Columns: 14
Preprocessing data ...
AutoGluon infers your prediction problem is: 'multiclass' (because dtype of label-column == object).
    First 10 (of 15) unique label values:  [' Exec-managerial', ' Other-service', ' Craft-repair', ' Sales', ' Prof-specialty', ' Protective-serv', ' ?', ' Adm-clerical', ' Machine-op-inspct', ' Tech-support']
    If 'multiclass' 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'])
Warning: Some classes in the training set have fewer than 10 examples. AutoGluon will only keep 12 out of 15 classes for training and will not try to predict the rare classes. To keep more classes, increase the number of datapoints from these rare classes in the training data or reduce label_count_threshold.
Fraction of data from classes with at least 10 examples that will be kept for training models: 0.978
Train Data Class Count: 12
Using Feature Generators to preprocess the data ...
Fitting AutoMLPipelineFeatureGenerator...
    Available Memory:                    22321.63 MB
    Train Data (Original)  Memory Usage: 3.11 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', 'relationship', 'race', ...]
    Types of features in processed data (raw dtype, special dtypes):
            ('category', []) : 8 | ['workclass', 'education', 'marital-status', 'relationship', 'race', ...]
            ('int', [])      : 6 | ['age', 'fnlwgt', 'education-num', 'capital-gain', 'capital-loss', ...]
    5.5s = Fit runtime
    14 features in original data used to generate 14 features in processed data.
    Train Data (Processed) Memory Usage: 0.3 MB (0.0% of available memory)
Data preprocessing and feature engineering runtime = 5.5s ...
AutoGluon will gauge predictive performance using evaluation metric: 'accuracy'
    To change this, specify the eval_metric argument of fit()
Fitting 2 L1 models ...
Hyperparameter tuning model: LightGBM ...
/var/lib/jenkins/workspace/workspace/autogluon-tutorial-tabular-v3/venv/lib/python3.7/site-packages/fsspec/__init__.py:49: DeprecationWarning: SelectableGroups dict interface is deprecated. Use select.
  for spec in entry_points.get("fsspec.specs", []):
  0%|          | 0/5 [00:00<?, ?it/s]
Fitted model: LightGBM/T0 ...
    0.3033   = Validation accuracy score
    0.66s    = Training runtime
    0.05s    = Validation runtime
Fitted model: LightGBM/T1 ...
    0.3065   = Validation accuracy score
    0.63s    = Training runtime
    0.06s    = Validation runtime
Fitted model: LightGBM/T2 ...
    0.3061   = Validation accuracy score
    0.59s    = Training runtime
    0.02s    = Validation runtime
Fitted model: LightGBM/T3 ...
    0.3074   = Validation accuracy score
    0.6s     = Training runtime
    0.08s    = Validation runtime
Fitted model: LightGBM/T4 ...
    0.2955   = Validation accuracy score
    0.6s     = Training runtime
    0.07s    = Validation runtime
Hyperparameter tuning model: NeuralNetMXNet ...
  0%|          | 0/5 [00:00<?, ?it/s]
Fitted model: NeuralNetMXNet/T0 ...
    0.1708   = Validation accuracy score
    5.31s    = Training runtime
    0.43s    = Validation runtime
Fitted model: NeuralNetMXNet/T1 ...
    0.1718   = Validation accuracy score
    5.2s     = Training runtime
    0.43s    = Validation runtime
Fitted model: NeuralNetMXNet/T2 ...
    0.1306   = Validation accuracy score
    5.19s    = Training runtime
    0.44s    = Validation runtime
Fitted model: NeuralNetMXNet/T3 ...
    0.1726   = Validation accuracy score
    5.72s    = Training runtime
    0.43s    = Validation runtime
Fitted model: NeuralNetMXNet/T4 ...
    0.2653   = Validation accuracy score
    5.46s    = Training runtime
    0.43s    = Validation runtime
Fitting model: WeightedEnsemble_L2 ... Training model for up to 114.5s of the 76.67s of remaining time.
    0.3119   = Validation accuracy score
    1.38s    = Training runtime
    0.0s     = Validation runtime
AutoGluon training complete, total runtime = 44.74s ...
TabularPredictor saved. To load, use: predictor = TabularPredictor.load("AutogluonModels/ag-20210617_032450/")

We again demonstrate how to use the trained models to predict on the test data.

y_pred = predictor.predict(test_data_nolabel)
print("Predictions:  ", list(y_pred)[:5])
perf = predictor.evaluate(test_data, auxiliary_metrics=False)
Predictions:   [' Exec-managerial', ' Craft-repair', ' Craft-repair', ' Other-service', ' Sales']
Evaluation: accuracy on test data: 0.28748165233801637
Evaluations on test data:
{
    "accuracy": 0.28748165233801637
}

Use the following to view a summary of what happened during fit. Now this command will show details of the hyperparameter-tuning process for each type of model:

results = predictor.fit_summary()
* Summary of fit() *
Estimated performance of each model:
                  model  score_val  pred_time_val   fit_time  pred_time_val_marginal  fit_time_marginal  stack_level  can_infer  fit_order
0   WeightedEnsemble_L2   0.311872       1.564228  20.436882                0.001187           1.384421            2       True         11
1           LightGBM/T3   0.307361       0.079673   0.596876                0.079673           0.596876            1       True          4
2           LightGBM/T1   0.306541       0.056236   0.628748                0.056236           0.628748            1       True          2
3           LightGBM/T2   0.306131       0.018912   0.593100                0.018912           0.593100            1       True          3
4           LightGBM/T0   0.303260       0.047034   0.661718                0.047034           0.661718            1       True          1
5           LightGBM/T4   0.295469       0.069978   0.604495                0.069978           0.604495            1       True          5
6     NeuralNetMXNet/T4   0.265327       0.432762   5.457770                0.432762           5.457770            1       True         10
7     NeuralNetMXNet/T3   0.172647       0.430578   5.724425                0.430578           5.724425            1       True          9
8     NeuralNetMXNet/T1   0.171827       0.428288   5.197539                0.428288           5.197539            1       True          7
9     NeuralNetMXNet/T0   0.170802       0.430158   5.312216                0.430158           5.312216            1       True          6
10    NeuralNetMXNet/T2   0.130613       0.440019   5.193011                0.440019           5.193011            1       True          8
Number of models trained: 11
Types of models trained:
{'TabularNeuralNetModel', 'WeightedEnsembleModel', 'LGBModel'}
Bagging used: False
Multi-layer stack-ensembling used: False
Feature Metadata (Processed):
(raw dtype, special dtypes):
('category', []) : 8 | ['workclass', 'education', 'marital-status', 'relationship', 'race', ...]
('int', [])      : 6 | ['age', 'fnlwgt', 'education-num', 'capital-gain', 'capital-loss', ...]
* End of fit() summary *
/var/lib/jenkins/workspace/workspace/autogluon-tutorial-tabular-v3/core/src/autogluon/core/utils/plots.py:140: UserWarning: AutoGluon summary plots cannot be created because bokeh is not installed. To see plots, please do: "pip install bokeh==2.0.1"
  warnings.warn('AutoGluon summary plots cannot be created because bokeh is not installed. To see plots, please do: "pip install bokeh==2.0.1"')

In the above example, the predictive performance may be poor because we specified very little training to ensure quick runtimes. You can call fit() multiple times while modifying the above settings to better understand how these choices affect performance outcomes. For example: you can comment out the train_data.head command or increase subsample_size to train using a larger dataset, increase the num_epochs and num_boost_round hyperparameters, and increase the time_limit (which you should do for all code in these tutorials). To see more detailed output during the execution of fit(), you can also pass in the argument: verbosity = 3.

Model ensembling with stacking/bagging

Beyond hyperparameter-tuning with a correctly-specified evaluation metric, two other methods to boost predictive performance are bagging and stack-ensembling. You’ll often see performance improve if you specify num_bag_folds = 5-10, num_stack_levels = 1-3 in the call to fit(), but this will increase training times and memory/disk usage.

predictor = TabularPredictor(label=label, eval_metric=metric).fit(train_data,
    num_bag_folds=5, num_bag_sets=1, num_stack_levels=1,
    hyperparameters = {'NN': {'num_epochs': 2}, 'GBM': {'num_boost_round': 20}},  # last  argument is just for quick demo here, omit it in real applications
)
No path specified. Models will be saved in: "AutogluonModels/ag-20210617_032539/"
Beginning AutoGluon training ...
AutoGluon will save models to "AutogluonModels/ag-20210617_032539/"
AutoGluon Version:  0.2.1b20210617
Train Data Rows:    500
Train Data Columns: 14
Preprocessing data ...
AutoGluon infers your prediction problem is: 'multiclass' (because dtype of label-column == object).
    First 10 (of 15) unique label values:  [' Exec-managerial', ' Other-service', ' Craft-repair', ' Sales', ' Prof-specialty', ' Protective-serv', ' ?', ' Adm-clerical', ' Machine-op-inspct', ' Tech-support']
    If 'multiclass' 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'])
Warning: Some classes in the training set have fewer than 10 examples. AutoGluon will only keep 12 out of 15 classes for training and will not try to predict the rare classes. To keep more classes, increase the number of datapoints from these rare classes in the training data or reduce label_count_threshold.
Fraction of data from classes with at least 10 examples that will be kept for training models: 0.978
Train Data Class Count: 12
Using Feature Generators to preprocess the data ...
Fitting AutoMLPipelineFeatureGenerator...
    Available Memory:                    22075.5 MB
    Train Data (Original)  Memory Usage: 0.28 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', 'relationship', 'race', ...]
    Types of features in processed data (raw dtype, special dtypes):
            ('category', []) : 8 | ['workclass', 'education', 'marital-status', 'relationship', 'race', ...]
            ('int', [])      : 6 | ['age', 'fnlwgt', 'education-num', 'capital-gain', 'capital-loss', ...]
    0.6s = 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 = 0.58s ...
AutoGluon will gauge predictive performance using evaluation metric: 'accuracy'
    To change this, specify the eval_metric argument of fit()
AutoGluon will fit 2 stack levels (L1 to L2) ...
Fitting 2 L1 models ...
Fitting model: LightGBM_BAG_L1 ...
    0.3067   = Validation accuracy score
    0.79s    = Training runtime
    0.05s    = Validation runtime
Fitting model: NeuralNetMXNet_BAG_L1 ...
    0.1002   = Validation accuracy score
    1.62s    = Training runtime
    0.12s    = Validation runtime
Fitting model: WeightedEnsemble_L2 ...
    0.3067   = Validation accuracy score
    0.12s    = Training runtime
    0.0s     = Validation runtime
Fitting 2 L2 models ...
Fitting model: LightGBM_BAG_L2 ...
    0.2924   = Validation accuracy score
    1.03s    = Training runtime
    0.05s    = Validation runtime
Fitting model: NeuralNetMXNet_BAG_L2 ...
    0.0818   = Validation accuracy score
    2.07s    = Training runtime
    0.16s    = Validation runtime
Fitting model: WeightedEnsemble_L3 ...
    0.2924   = Validation accuracy score
    0.12s    = Training runtime
    0.0s     = Validation runtime
AutoGluon training complete, total runtime = 6.85s ...
TabularPredictor saved. To load, use: predictor = TabularPredictor.load("AutogluonModels/ag-20210617_032539/")

You should not provide tuning_data when stacking/bagging, and instead provide all your available data as train_data (which AutoGluon will split in more intellgent ways). num_bag_sets controls how many times the k-fold bagging process is repeated to further reduce variance (increasing this may further boost accuracy but will substantially increase training times, inference latency, and memory/disk usage). Rather than manually searching for good bagging/stacking values yourself, AutoGluon will automatically select good values for you if you specify auto_stack instead:

save_path = 'agModels-predictOccupation'  # folder where to store trained models

predictor = TabularPredictor(label=label, eval_metric=metric, path=save_path).fit(
    train_data, auto_stack=True,
    time_limit=30, hyperparameters={'NN': {'num_epochs': 2}, 'GBM': {'num_boost_round': 20}}  # last 2 arguments are for quick demo, omit them in real applications
)
Beginning AutoGluon training ... Time limit = 30s
AutoGluon will save models to "agModels-predictOccupation/"
AutoGluon Version:  0.2.1b20210617
Train Data Rows:    500
Train Data Columns: 14
Preprocessing data ...
AutoGluon infers your prediction problem is: 'multiclass' (because dtype of label-column == object).
    First 10 (of 15) unique label values:  [' Exec-managerial', ' Other-service', ' Craft-repair', ' Sales', ' Prof-specialty', ' Protective-serv', ' ?', ' Adm-clerical', ' Machine-op-inspct', ' Tech-support']
    If 'multiclass' 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'])
Warning: Some classes in the training set have fewer than 10 examples. AutoGluon will only keep 12 out of 15 classes for training and will not try to predict the rare classes. To keep more classes, increase the number of datapoints from these rare classes in the training data or reduce label_count_threshold.
Fraction of data from classes with at least 10 examples that will be kept for training models: 0.978
Train Data Class Count: 12
Using Feature Generators to preprocess the data ...
Fitting AutoMLPipelineFeatureGenerator...
    Available Memory:                    22050.77 MB
    Train Data (Original)  Memory Usage: 0.28 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', 'relationship', 'race', ...]
    Types of features in processed data (raw dtype, special dtypes):
            ('category', []) : 8 | ['workclass', 'education', 'marital-status', 'relationship', 'race', ...]
            ('int', [])      : 6 | ['age', 'fnlwgt', 'education-num', 'capital-gain', 'capital-loss', ...]
    0.6s = 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 = 0.58s ...
AutoGluon will gauge predictive performance using evaluation metric: 'accuracy'
    To change this, specify the eval_metric argument of fit()
Fitting 2 L1 models ...
Fitting model: LightGBM_BAG_L1 ... Training model for up to 29.42s of the 29.42s of remaining time.
    0.3067   = Validation accuracy score
    0.83s    = Training runtime
    0.05s    = Validation runtime
Fitting model: NeuralNetMXNet_BAG_L1 ... Training model for up to 28.52s of the 28.52s of remaining time.
    0.0941   = Validation accuracy score
    1.95s    = Training runtime
    0.12s    = Validation runtime
Repeating k-fold bagging: 2/20
Fitting model: LightGBM_BAG_L1 ... Training model for up to 26.41s of the 26.41s of remaining time.
    0.3149   = Validation accuracy score
    1.64s    = Training runtime
    0.09s    = Validation runtime
Fitting model: NeuralNetMXNet_BAG_L1 ... Training model for up to 25.53s of the 25.53s of remaining time.
    0.0859   = Validation accuracy score
    3.91s    = Training runtime
    0.24s    = Validation runtime
Repeating k-fold bagging: 3/20
Fitting model: LightGBM_BAG_L1 ... Training model for up to 23.42s of the 23.42s of remaining time.
    0.3292   = Validation accuracy score
    2.47s    = Training runtime
    0.14s    = Validation runtime
Fitting model: NeuralNetMXNet_BAG_L1 ... Training model for up to 22.52s of the 22.52s of remaining time.
    0.1084   = Validation accuracy score
    5.76s    = Training runtime
    0.35s    = Validation runtime
Repeating k-fold bagging: 4/20
Fitting model: LightGBM_BAG_L1 ... Training model for up to 20.51s of the 20.51s of remaining time.
    0.3108   = Validation accuracy score
    3.26s    = Training runtime
    0.19s    = Validation runtime
Fitting model: NeuralNetMXNet_BAG_L1 ... Training model for up to 19.65s of the 19.64s of remaining time.
    0.1002   = Validation accuracy score
    7.72s    = Training runtime
    0.47s    = Validation runtime
Repeating k-fold bagging: 5/20
Fitting model: LightGBM_BAG_L1 ... Training model for up to 17.53s of the 17.53s of remaining time.
    0.3129   = Validation accuracy score
    4.07s    = Training runtime
    0.23s    = Validation runtime
Fitting model: NeuralNetMXNet_BAG_L1 ... Training model for up to 16.66s of the 16.66s of remaining time.
    0.1002   = Validation accuracy score
    9.7s     = Training runtime
    0.59s    = Validation runtime
Repeating k-fold bagging: 6/20
Fitting model: LightGBM_BAG_L1 ... Training model for up to 14.51s of the 14.51s of remaining time.
    0.3108   = Validation accuracy score
    4.87s    = Training runtime
    0.28s    = Validation runtime
Fitting model: NeuralNetMXNet_BAG_L1 ... Training model for up to 13.65s of the 13.65s of remaining time.
    0.0941   = Validation accuracy score
    11.55s   = Training runtime
    0.71s    = Validation runtime
Repeating k-fold bagging: 7/20
Fitting model: LightGBM_BAG_L1 ... Training model for up to 11.64s of the 11.64s of remaining time.
    0.3088   = Validation accuracy score
    5.71s    = Training runtime
    0.33s    = Validation runtime
Fitting model: NeuralNetMXNet_BAG_L1 ... Training model for up to 10.72s of the 10.71s of remaining time.
    0.1002   = Validation accuracy score
    13.51s   = Training runtime
    0.83s    = Validation runtime
Repeating k-fold bagging: 8/20
Fitting model: LightGBM_BAG_L1 ... Training model for up to 8.59s of the 8.59s of remaining time.
    0.3067   = Validation accuracy score
    6.53s    = Training runtime
    0.38s    = Validation runtime
Fitting model: NeuralNetMXNet_BAG_L1 ... Training model for up to 7.7s of the 7.7s of remaining time.
    0.137    = Validation accuracy score
    15.48s   = Training runtime
    0.94s    = Validation runtime
Repeating k-fold bagging: 9/20
Fitting model: LightGBM_BAG_L1 ... Training model for up to 5.58s of the 5.58s of remaining time.
    0.3047   = Validation accuracy score
    7.32s    = Training runtime
    0.42s    = Validation runtime
Fitting model: NeuralNetMXNet_BAG_L1 ... Training model for up to 4.71s of the 4.71s of remaining time.
    0.1309   = Validation accuracy score
    17.33s   = Training runtime
    1.06s    = Validation runtime
Completed 9/20 k-fold bagging repeats ...
Fitting model: WeightedEnsemble_L2 ... Training model for up to 29.42s of the 2.69s of remaining time.
    0.3088   = Validation accuracy score
    0.12s    = Training runtime
    0.0s     = Validation runtime
AutoGluon training complete, total runtime = 27.44s ...
TabularPredictor saved. To load, use: predictor = TabularPredictor.load("agModels-predictOccupation/")

Often stacking/bagging will produce superior accuracy than hyperparameter-tuning, but you may try combining both techniques (note: specifying presets='best_quality' in fit() simply sets auto_stack=True).

Prediction options (inference)

Even if you’ve started a new Python session since last calling fit(), you can still load a previously trained predictor from disk:

predictor = TabularPredictor.load(save_path)  # `predictor.path` is another way to get the relative path needed to later load predictor.

Above save_path is the same folder previously passed to TabularPredictor, in which all the trained models have been saved. You can train easily models on one machine and deploy them on another. Simply copy the save_path folder to the new machine and specify its new path in TabularPredictor.load().

We can make a prediction on an individual example rather than a full dataset:

datapoint = test_data_nolabel.iloc[[0]]  # Note: .iloc[0] won't work because it returns pandas Series instead of DataFrame
print(datapoint)
predictor.predict(datapoint)
      age workclass  fnlwgt      education  education-num marital-status  5000   49   Private  259087   Some-college             10       Divorced

        relationship    race      sex  capital-gain  capital-loss  5000   Not-in-family   White   Female             0             0

      hours-per-week  native-country   class
5000              40   United-States   <=50K
5000     Exec-managerial
Name: occupation, dtype: object

To output predicted class probabilities instead of predicted classes, you can use:

predictor.predict_proba(datapoint)  # returns a DataFrame that shows which probability corresponds to which class
? Adm-clerical Armed-Forces Craft-repair Exec-managerial Farming-fishing Handlers-cleaners Machine-op-inspct Other-service Priv-house-serv Prof-specialty Protective-serv Sales Tech-support Transport-moving
5000 0.048792 0.139282 0.0 0.123637 0.187524 0.031939 0.048805 0.058896 0.067417 0.0 0.089078 0.0 0.100272 0.032965 0.071394

By default, predict() and predict_proba() will utilize the model that AutoGluon thinks is most accurate, which is usually an ensemble of many individual models. Here’s how to see which model this is:

predictor.get_model_best()
'WeightedEnsemble_L2'

We can instead specify a particular model to use for predictions (e.g. to reduce inference latency). Note that a ‘model’ in AutoGluon may refer to for example a single Neural Network, a bagged ensemble of many Neural Network copies trained on different training/validation splits, a weighted ensemble that aggregates the predictions of many other models, or a stacker model that operates on predictions output by other models. This is akin to viewing a Random Forest as one ‘model’ when it is in fact an ensemble of many decision trees.

Before deciding which model to use, let’s evaluate all of the models AutoGluon has previously trained on our test data:

predictor.leaderboard(test_data, silent=True)
model score_test score_val pred_time_test pred_time_val fit_time pred_time_test_marginal pred_time_val_marginal fit_time_marginal stack_level can_infer fit_order
0 LightGBM_BAG_L1 0.282449 0.304703 0.908189 0.421562 7.323407 0.908189 0.421562 7.323407 1 True 1
1 WeightedEnsemble_L2 0.280562 0.308793 20.035225 1.483705 24.772502 0.002649 0.000479 0.117128 2 True 3
2 NeuralNetMXNet_BAG_L1 0.129377 0.130879 19.124388 1.061664 17.331967 19.124388 1.061664 17.331967 1 True 2

The leaderboard shows each model’s predictive performance on the test data (score_test) and validation data (score_val), as well as the time required to: produce predictions for the test data (pred_time_val), produce predictions on the validation data (pred_time_val), and train only this model (fit_time). Below, we show that a leaderboard can be produced without new data (just uses the data previously reserved for validation inside fit) and can display extra information about each model:

predictor.leaderboard(extra_info=True, silent=True)
model score_val pred_time_val fit_time pred_time_val_marginal fit_time_marginal stack_level can_infer fit_order num_features ... child_model_type hyperparameters hyperparameters_fit ag_args_fit features child_hyperparameters child_hyperparameters_fit child_ag_args_fit ancestors descendants
0 WeightedEnsemble_L2 0.308793 1.483705 24.772502 0.000479 0.117128 2 True 3 24 ... GreedyWeightedEnsembleModel {'use_orig_features': False, 'max_base_models'... {} {'max_memory_usage_ratio': 1.0, 'max_time_limi... [LightGBM_BAG_L1_4, LightGBM_BAG_L1_11, LightG... {'ensemble_size': 100} {'ensemble_size': 4} {'max_memory_usage_ratio': 1.0, 'max_time_limi... [NeuralNetMXNet_BAG_L1, LightGBM_BAG_L1] []
1 LightGBM_BAG_L1 0.304703 0.421562 7.323407 0.421562 7.323407 1 True 1 14 ... LGBModel {'use_orig_features': True, 'max_base_models':... {} {'max_memory_usage_ratio': 1.0, 'max_time_limi... [marital-status, sex, workclass, relationship,... {'num_boost_round': 20, 'num_threads': -1, 'le... {'num_boost_round': 11} {'max_memory_usage_ratio': 1.0, 'max_time_limi... [] [WeightedEnsemble_L2]
2 NeuralNetMXNet_BAG_L1 0.130879 1.061664 17.331967 1.061664 17.331967 1 True 2 14 ... TabularNeuralNetModel {'use_orig_features': True, 'max_base_models':... {} {'max_memory_usage_ratio': 1.0, 'max_time_limi... [marital-status, sex, workclass, relationship,... {'num_epochs': 2, 'epochs_wo_improve': 20, 'se... {'num_epochs': 2} {'max_memory_usage_ratio': 1.0, 'max_time_limi... [] [WeightedEnsemble_L2]

3 rows × 29 columns

The expanded leaderboard shows properties like how many features are used by each model (num_features), which other models are ancestors whose predictions are required inputs for each model (ancestors), and how much memory each model and all its ancestors would occupy if simultaneously persisted (memory_size_w_ancestors). See the leaderboard documentation for full details.

To show scores for other metrics, you can specify the extra_metrics argument when passing in test_data:

predictor.leaderboard(test_data, extra_metrics=['accuracy', 'balanced_accuracy', 'log_loss'], silent=True)
model score_test accuracy balanced_accuracy log_loss score_val pred_time_test pred_time_val fit_time pred_time_test_marginal pred_time_val_marginal fit_time_marginal stack_level can_infer fit_order
0 LightGBM_BAG_L1 0.282449 0.282449 0.176065 -11.747265 0.304703 0.917826 0.421562 7.323407 0.917826 0.421562 7.323407 1 True 1
1 WeightedEnsemble_L2 0.280562 0.280562 0.174767 -11.726069 0.308793 20.157671 1.483705 24.772502 0.002666 0.000479 0.117128 2 True 3
2 NeuralNetMXNet_BAG_L1 0.129377 0.129377 0.067970 -11.765665 0.130879 19.237179 1.061664 17.331967 19.237179 1.061664 17.331967 1 True 2

Notice that log_loss scores are negative. This is because metrics in AutoGluon are always shown in higher_is_better form. This means that metrics such as log_loss and root_mean_squared_error will have their signs FLIPPED, and values will be negative. This is necessary to avoid the user needing to know the metric to understand if higher is better when looking at leaderboard.

One additional caviat: It is possible that log_loss values can be -inf when computed via extra_metrics. This is because the models were not optimized with log_loss in mind during training and may have prediction probabilities giving a class 0 (particularly common with K-Nearest-Neighbors models). Because log_loss gives infinite error when the correct class was given 0 probability, this results in a score of -inf. It is therefore recommended that log_loss should not be used as a secondary metric to determine model quality. Either use log_loss as the eval_metric or avoid it altogether.

Here’s how to specify a particular model to use for prediction instead of AutoGluon’s default model-choice:

i = 0  # index of model to use
model_to_use = predictor.get_model_names()[i]
model_pred = predictor.predict(datapoint, model=model_to_use)
print("Prediction from %s model: %s" % (model_to_use, model_pred.iloc[0]))
Prediction from LightGBM_BAG_L1 model:  Exec-managerial

We can easily access various information about the trained predictor or a particular model:

all_models = predictor.get_model_names()
model_to_use = all_models[i]
specific_model = predictor._trainer.load_model(model_to_use)

# Objects defined below are dicts of various information (not printed here as they are quite large):
model_info = specific_model.get_info()
predictor_information = predictor.info()

The predictor also remembers what metric predictions should be evaluated with, which can be done with ground truth labels as follows:

y_pred_proba = predictor.predict_proba(test_data_nolabel)
perf = predictor.evaluate_predictions(y_true=y_test, y_pred=y_pred_proba)
Evaluation: accuracy on test data: 0.28056196267561334
Evaluations on test data:
{
    "accuracy": 0.28056196267561334,
    "balanced_accuracy": 0.17476705377211224,
    "mcc": 0.18854859974947066
}

Since the label columns remains in the test_data DataFrame, we can instead use the shorthand:

perf = predictor.evaluate(test_data)
Evaluation: accuracy on test data: 0.28056196267561334
Evaluations on test data:
{
    "accuracy": 0.28056196267561334,
    "balanced_accuracy": 0.17476705377211224,
    "mcc": 0.18854859974947066
}

Interpretability (feature importance)

To better understand our trained predictor, we can estimate the overall importance of each feature:

predictor.feature_importance(test_data)
Computing feature importance via permutation shuffling for 14 features using 1000 rows with 3 shuffle sets...
    262.23s = Expected runtime (87.41s per shuffle set)
    191.56s = Actual runtime (Completed 3 of 3 shuffle sets)
importance stddev p_value n p99_high p99_low
education-num 7.100000e-02 0.007550 0.001874 3 0.114261 0.027739
workclass 4.133333e-02 0.004726 0.002165 3 0.068413 0.014254
sex 3.566667e-02 0.014572 0.025694 3 0.119164 -0.047831
hours-per-week 2.700000e-02 0.005196 0.006061 3 0.056775 -0.002775
age 1.666667e-02 0.010693 0.057088 3 0.077937 -0.044604
class 5.333333e-03 0.006506 0.145754 3 0.042616 -0.031949
fnlwgt 2.666667e-03 0.016503 0.402929 3 0.097228 -0.091895
education 6.666667e-04 0.002517 0.345697 3 0.015087 -0.013754
race 0.000000e+00 0.000000 0.500000 3 0.000000 0.000000
native-country 0.000000e+00 0.000000 0.500000 3 0.000000 0.000000
marital-status -1.850372e-17 0.003464 0.500000 3 0.019850 -0.019850
capital-gain -6.666667e-04 0.000577 0.908248 3 0.002642 -0.003975
capital-loss -6.666667e-04 0.000577 0.908248 3 0.002642 -0.003975
relationship -1.666667e-03 0.002517 0.814970 3 0.012754 -0.016087

Computed via permutation-shuffling, these feature importance scores quantify the drop in predictive performance (of the already trained predictor) when one column’s values are randomly shuffled across rows. The top features in this list contribute most to AutoGluon’s accuracy (for predicting when/if a patient will be readmitted to the hospital). Features with non-positive importance score hardly contribute to the predictor’s accuracy, or may even be actively harmful to include in the data (consider removing these features from your data and calling fit again). These scores facilitate interpretability of the predictor’s global behavior (which features it relies on for all predictions) rather than local explanations that only rationalize one particular prediction.

Accelerating inference

We describe multiple ways to reduce the time it takes for AutoGluon to produce predictions.

Keeping models in memory

By default, AutoGluon loads models into memory one at a time and only when they are needed for prediction. This strategy is robust for large stacked/bagged ensembles, but leads to slower prediction times. If you plan to repeatedly make predictions (e.g. on new datapoints one at a time rather than one large test dataset), you can first specify that all models required for inference should be loaded into memory as follows:

predictor.persist_models()

num_test = 20
preds = np.array(['']*num_test, dtype='object')
for i in range(num_test):
    datapoint = test_data_nolabel.iloc[[i]]
    pred_numpy = predictor.predict(datapoint, as_pandas=False)
    preds[i] = pred_numpy[0]

perf = predictor.evaluate_predictions(y_test[:num_test], preds, auxiliary_metrics=True)
print("Predictions: ", preds)

predictor.unpersist_models()  # free memory by clearing models, future predict() calls will load models from disk
Persisting 3 models in memory. Models will require 0.22% of memory.
Evaluation: accuracy on test data: 0.25
Evaluations on test data:
{
    "accuracy": 0.25,
    "balanced_accuracy": 0.3208333333333336,
    "mcc": 0.12121844240150972
}
Unpersisted 3 models: ['NeuralNetMXNet_BAG_L1', 'LightGBM_BAG_L1', 'WeightedEnsemble_L2']
Predictions:  [' Exec-managerial' ' Craft-repair' ' Craft-repair' ' ?' ' ?'
 ' Exec-managerial' ' Exec-managerial' ' Sales' ' Exec-managerial'
 ' Adm-clerical' ' Other-service' ' Exec-managerial' ' Exec-managerial'
 ' Exec-managerial' ' Adm-clerical' ' ?' ' Craft-repair' ' Craft-repair'
 ' Exec-managerial' ' Craft-repair']
['NeuralNetMXNet_BAG_L1', 'LightGBM_BAG_L1', 'WeightedEnsemble_L2']

You can alternatively specify a particular model to persist via the models argument of persist_models(), or simply set models='all' to simultaneously load every single model that was trained during fit.

Using smaller ensemble or faster model for prediction

Without having to retrain any models, one can construct alternative ensembles that aggregate individual models’ predictions with different weighting schemes. These ensembles become smaller (and hence faster for prediction) if they assign nonzero weight to less models. You can produce a wide variety of ensembles with different accuracy-speed tradeoffs like this:

additional_ensembles = predictor.fit_weighted_ensemble(expand_pareto_frontier=True)
print("Alternative ensembles you can use for prediction:", additional_ensembles)

predictor.leaderboard(only_pareto_frontier=True, silent=True)
Fitting model: WeightedEnsemble_L2Best ...
    0.3088   = Validation accuracy score
    0.12s    = Training runtime
    0.0s     = Validation runtime
Alternative ensembles you can use for prediction: ['WeightedEnsemble_L2Best']
model score_val pred_time_val fit_time pred_time_val_marginal fit_time_marginal stack_level can_infer fit_order
0 WeightedEnsemble_L2 0.308793 1.483705 24.772502 0.000479 0.117128 2 True 3
1 LightGBM_BAG_L1 0.304703 0.421562 7.323407 0.421562 7.323407 1 True 1

The resulting leaderboard will contain the most accurate model for a given inference-latency. You can select whichever model exhibits acceptable latency from the leaderboard and use it for prediction.

model_for_prediction = additional_ensembles[0]
predictions = predictor.predict(test_data, model=model_for_prediction)
predictor.delete_models(models_to_delete=additional_ensembles, dry_run=False)  # delete these extra models so they don't affect rest of tutorial
Deleting model WeightedEnsemble_L2Best. All files under agModels-predictOccupation/models/WeightedEnsemble_L2Best/ will be removed.

Collapsing bagged ensembles via refit_full

For an ensemble predictor trained with bagging (as done above), recall there ~10 bagged copies of each individual model trained on different train/validation folds. We can collapse this bag of ~10 models into a single model that’s fit to the full dataset, which can greatly reduce its memory/latency requirements (but may also reduce accuracy). Below we refit such a model for each original model but you can alternatively do this for just a particular model by specifying the model argument of refit_full().

refit_model_map = predictor.refit_full()
print("Name of each refit-full model corresponding to a previous bagged ensemble:")
print(refit_model_map)
predictor.leaderboard(test_data, silent=True)
Fitting 1 L1 models ...
Fitting model: LightGBM_BAG_L1_FULL ...
    0.17s    = Training runtime
Fitting 1 L1 models ...
Fitting model: NeuralNetMXNet_BAG_L1_FULL ...
    0.31s    = Training runtime
Fitting model: WeightedEnsemble_L2_FULL ...
    0.3088   = Validation accuracy score
    0.01s    = Training runtime
    0.0s     = Validation runtime
Name of each refit-full model corresponding to a previous bagged ensemble:
{'LightGBM_BAG_L1': 'LightGBM_BAG_L1_FULL', 'NeuralNetMXNet_BAG_L1': 'NeuralNetMXNet_BAG_L1_FULL', 'WeightedEnsemble_L2': 'WeightedEnsemble_L2_FULL'}
model score_test score_val pred_time_test pred_time_val fit_time pred_time_test_marginal pred_time_val_marginal fit_time_marginal stack_level can_infer fit_order
0 LightGBM_BAG_L1 0.282449 0.304703 0.910169 0.421562 7.323407 0.910169 0.421562 7.323407 1 True 1
1 LightGBM_BAG_L1_FULL 0.281191 NaN 0.032001 NaN 0.170923 0.032001 NaN 0.170923 1 True 4
2 WeightedEnsemble_L2_FULL 0.280981 NaN 0.491991 NaN 0.481479 0.002369 0.000521 0.005110 2 True 6
3 WeightedEnsemble_L2 0.280562 0.308793 20.153367 1.483705 24.772502 0.002733 0.000479 0.117128 2 True 3
4 NeuralNetMXNet_BAG_L1 0.129377 0.130879 19.240464 1.061664 17.331967 19.240464 1.061664 17.331967 1 True 2
5 NeuralNetMXNet_BAG_L1_FULL 0.112602 NaN 0.457621 NaN 0.305446 0.457621 NaN 0.305446 1 True 5

This adds the refit-full models to the leaderboard and we can opt to use any of them for prediction just like any other model. Note pred_time_test and pred_time_val list the time taken to produce predictions with each model (in seconds) on the test/validation data. Since the refit-full models were trained using all of the data, there is no internal validation score (score_val) available for them. You can also call refit_full() with non-bagged models to refit the same models to your full dataset (there won’t be memory/latency gains in this case but test accuracy may improve).

Model distillation

While computationally-favorable, single individual models will usually have lower accuracy than weighted/stacked/bagged ensembles. Model Distillation offers one way to retain the computational benefits of a single model, while enjoying some of the accuracy-boost that comes with ensembling. The idea is to train the individual model (which we can call the student) to mimic the predictions of the full stack ensemble (the teacher). Like refit_full(), the distill() function will produce additional models we can opt to use for prediction.

student_models = predictor.distill(time_limit=30)  # specify much longer time limit in real applications
print(student_models)
preds_student = predictor.predict(test_data_nolabel, model=student_models[0])
print(f"predictions from {student_models[0]}:", list(preds_student)[:5])
predictor.leaderboard(test_data)
Distilling with teacher='WeightedEnsemble_L2', teacher_preds=soft, augment_method=spunge ...
SPUNGE: Augmenting training data with 1955 synthetic samples for distillation...
Distilling with each of these student models: ['LightGBM_DSTL', 'NeuralNetMXNet_DSTL', 'RandomForestMSE_DSTL', 'CatBoost_DSTL']
Fitting 4 L1 models ...
Fitting model: LightGBM_DSTL ... Training model for up to 30.0s of the 30.0s of remaining time.
    Note: model has different eval_metric than default.
    -1.9746  = Validation soft_log_loss score
    4.16s    = Training runtime
    0.02s    = Validation runtime
Fitting model: NeuralNetMXNet_DSTL ... Training model for up to 25.69s of the 25.69s of remaining time.
    Note: model has different eval_metric than default.
    -2.0557  = Validation soft_log_loss score
    11.09s   = Training runtime
    0.02s    = Validation runtime
Fitting model: RandomForestMSE_DSTL ... Training model for up to 14.57s of the 14.57s of remaining time.
    Note: model has different eval_metric than default.
    -2.0449  = Validation soft_log_loss score
    0.92s    = Training runtime
    0.11s    = Validation runtime
Fitting model: CatBoost_DSTL ... Training model for up to 13.4s of the 13.4s of remaining time.
    Warning: Exception caused CatBoost_DSTL to fail during training (ImportError)... Skipping this model.
            import catboost_dev failed (needed for distillation with CatBoost models). Make sure you can import catboost and then run: 'pip install catboost-dev'.Detailed info: No module named 'catboost_dev'
Distilling with each of these student models: ['WeightedEnsemble_L2_DSTL']
Fitting model: WeightedEnsemble_L2_DSTL ... Training model for up to 30.0s of the 13.02s of remaining time.
    Note: model has different eval_metric than default.
    -1.9746  = Validation soft_log_loss score
    0.29s    = Training runtime
    0.0s     = Validation runtime
Distilled model leaderboard:
                      model  score_val  pred_time_val   fit_time  pred_time_val_marginal  fit_time_marginal  stack_level  can_infer  fit_order
0             LightGBM_DSTL   0.387755       0.018613   4.158749                0.018613           4.158749            1       True          7
1  WeightedEnsemble_L2_DSTL   0.387755       0.019767   4.445672                0.001154           0.286923            2       True         10
2       NeuralNetMXNet_DSTL   0.346939       0.023903  11.090570                0.023903          11.090570            1       True          8
3      RandomForestMSE_DSTL   0.306122       0.107957   0.921869                0.107957           0.921869            1       True          9
['LightGBM_DSTL', 'NeuralNetMXNet_DSTL', 'RandomForestMSE_DSTL', 'WeightedEnsemble_L2_DSTL']
predictions from LightGBM_DSTL: [' Exec-managerial', ' Exec-managerial', ' Craft-repair', ' Other-service', ' Craft-repair']
                        model  score_test  score_val  pred_time_test  pred_time_val   fit_time  pred_time_test_marginal  pred_time_val_marginal  fit_time_marginal  stack_level  can_infer  fit_order
0        RandomForestMSE_DSTL    0.296498   0.306122        0.246918       0.107957   0.921869                 0.246918                0.107957           0.921869            1       True          9
1         NeuralNetMXNet_DSTL    0.292304   0.346939        0.462026       0.023903  11.090570                 0.462026                0.023903          11.090570            1       True          8
2               LightGBM_DSTL    0.290208   0.387755        0.442984       0.018613   4.158749                 0.442984                0.018613           4.158749            1       True          7
3    WeightedEnsemble_L2_DSTL    0.290208   0.387755        0.444733       0.019767   4.445672                 0.001750                0.001154           0.286923            2       True         10
4             LightGBM_BAG_L1    0.282449   0.304703        0.935912       0.421562   7.323407                 0.935912                0.421562           7.323407            1       True          1
5        LightGBM_BAG_L1_FULL    0.281191        NaN        0.032544            NaN   0.170923                 0.032544                     NaN           0.170923            1       True          4
6    WeightedEnsemble_L2_FULL    0.280981        NaN        0.486103            NaN   0.481479                 0.002271                0.000521           0.005110            2       True          6
7         WeightedEnsemble_L2    0.280562   0.308793       20.245016       1.483705  24.772502                 0.005140                0.000479           0.117128            2       True          3
8       NeuralNetMXNet_BAG_L1    0.129377   0.130879       19.303963       1.061664  17.331967                19.303963                1.061664          17.331967            1       True          2
9  NeuralNetMXNet_BAG_L1_FULL    0.112602        NaN        0.451288            NaN   0.305446                 0.451288                     NaN           0.305446            1       True          5
model score_test score_val pred_time_test pred_time_val fit_time pred_time_test_marginal pred_time_val_marginal fit_time_marginal stack_level can_infer fit_order
0 RandomForestMSE_DSTL 0.296498 0.306122 0.246918 0.107957 0.921869 0.246918 0.107957 0.921869 1 True 9
1 NeuralNetMXNet_DSTL 0.292304 0.346939 0.462026 0.023903 11.090570 0.462026 0.023903 11.090570 1 True 8
2 LightGBM_DSTL 0.290208 0.387755 0.442984 0.018613 4.158749 0.442984 0.018613 4.158749 1 True 7
3 WeightedEnsemble_L2_DSTL 0.290208 0.387755 0.444733 0.019767 4.445672 0.001750 0.001154 0.286923 2 True 10
4 LightGBM_BAG_L1 0.282449 0.304703 0.935912 0.421562 7.323407 0.935912 0.421562 7.323407 1 True 1
5 LightGBM_BAG_L1_FULL 0.281191 NaN 0.032544 NaN 0.170923 0.032544 NaN 0.170923 1 True 4
6 WeightedEnsemble_L2_FULL 0.280981 NaN 0.486103 NaN 0.481479 0.002271 0.000521 0.005110 2 True 6
7 WeightedEnsemble_L2 0.280562 0.308793 20.245016 1.483705 24.772502 0.005140 0.000479 0.117128 2 True 3
8 NeuralNetMXNet_BAG_L1 0.129377 0.130879 19.303963 1.061664 17.331967 19.303963 1.061664 17.331967 1 True 2
9 NeuralNetMXNet_BAG_L1_FULL 0.112602 NaN 0.451288 NaN 0.305446 0.451288 NaN 0.305446 1 True 5

Faster presets or hyperparameters

Instead of trying to speed up a cumbersome trained model at prediction time, if you know inference latency or memory will be an issue at the outset, then you can adjust the training process accordingly to ensure fit() does not produce unwieldy models.

One option is to specify more lightweight presets:

presets = ['good_quality_faster_inference_only_refit', 'optimize_for_deployment']
predictor_light = TabularPredictor(label=label, eval_metric=metric).fit(train_data, presets=presets, time_limit=30)
No path specified. Models will be saved in: "AutogluonModels/ag-20210617_033321/"
Presets specified: ['good_quality_faster_inference_only_refit', 'optimize_for_deployment']
Beginning AutoGluon training ... Time limit = 30s
AutoGluon will save models to "AutogluonModels/ag-20210617_033321/"
AutoGluon Version:  0.2.1b20210617
Train Data Rows:    500
Train Data Columns: 14
Preprocessing data ...
AutoGluon infers your prediction problem is: 'multiclass' (because dtype of label-column == object).
    First 10 (of 15) unique label values:  [' Exec-managerial', ' Other-service', ' Craft-repair', ' Sales', ' Prof-specialty', ' Protective-serv', ' ?', ' Adm-clerical', ' Machine-op-inspct', ' Tech-support']
    If 'multiclass' 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'])
Warning: Some classes in the training set have fewer than 10 examples. AutoGluon will only keep 12 out of 15 classes for training and will not try to predict the rare classes. To keep more classes, increase the number of datapoints from these rare classes in the training data or reduce label_count_threshold.
Fraction of data from classes with at least 10 examples that will be kept for training models: 0.978
Train Data Class Count: 12
Using Feature Generators to preprocess the data ...
Fitting AutoMLPipelineFeatureGenerator...
    Available Memory:                    21868.18 MB
    Train Data (Original)  Memory Usage: 0.28 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', 'relationship', 'race', ...]
    Types of features in processed data (raw dtype, special dtypes):
            ('category', []) : 8 | ['workclass', 'education', 'marital-status', 'relationship', 'race', ...]
            ('int', [])      : 6 | ['age', 'fnlwgt', 'education-num', 'capital-gain', 'capital-loss', ...]
    0.6s = 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 = 0.59s ...
AutoGluon will gauge predictive performance using evaluation metric: 'accuracy'
    To change this, specify the eval_metric argument of fit()
Fitting 11 L1 models ...
Fitting model: NeuralNetFastAI_BAG_L1 ... Training model for up to 29.41s of the 29.41s of remaining time.
/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:115.)
  return torch._C._cuda_getDeviceCount() > 0
No improvement since epoch 7: early stopping
No improvement since epoch 7: early stopping
No improvement since epoch 7: early stopping
    0.2802   = Validation accuracy score
    3.48s    = Training runtime
    0.08s    = Validation runtime
Fitting model: LightGBMXT_BAG_L1 ... Training model for up to 25.84s of the 25.84s of remaining time.
    0.3558   = Validation accuracy score
    3.29s    = Training runtime
    0.06s    = Validation runtime
Fitting model: LightGBM_BAG_L1 ... Training model for up to 22.48s of the 22.48s of remaining time.
    0.3395   = Validation accuracy score
    3.73s    = Training runtime
    0.05s    = Validation runtime
Fitting model: RandomForestGini_BAG_L1 ... Training model for up to 18.68s of the 18.68s of remaining time.
    0.3047   = Validation accuracy score
    0.71s    = Training runtime
    0.1s     = Validation runtime
Fitting model: RandomForestEntr_BAG_L1 ... Training model for up to 17.87s of the 17.87s of remaining time.
    0.2945   = Validation accuracy score
    0.61s    = Training runtime
    0.1s     = Validation runtime
Fitting model: CatBoost_BAG_L1 ... Training model for up to 17.15s of the 17.15s of remaining time.
    Time limit exceeded... Skipping CatBoost_BAG_L1.
Fitting model: ExtraTreesGini_BAG_L1 ... Training model for up to 13.15s of the 13.15s of remaining time.
    0.3252   = Validation accuracy score
    0.71s    = Training runtime
    0.1s     = Validation runtime
Fitting model: ExtraTreesEntr_BAG_L1 ... Training model for up to 12.34s of the 12.34s of remaining time.
    0.3088   = Validation accuracy score
    0.61s    = Training runtime
    0.1s     = Validation runtime
Fitting model: XGBoost_BAG_L1 ... Training model for up to 11.62s of the 11.62s of remaining time.
    0.3476   = Validation accuracy score
    3.66s    = Training runtime
    0.03s    = Validation runtime
Fitting model: NeuralNetMXNet_BAG_L1 ... Training model for up to 7.91s of the 7.91s of remaining time.
    Ran out of time, stopping training early. (Stopping on epoch 19)
    Ran out of time, stopping training early. (Stopping on epoch 20)
    Ran out of time, stopping training early. (Stopping on epoch 20)
    Ran out of time, stopping training early. (Stopping on epoch 23)
    Ran out of time, stopping training early. (Stopping on epoch 27)
    0.2168   = Validation accuracy score
    7.48s    = Training runtime
    0.12s    = Validation runtime
Fitting model: LightGBMLarge_BAG_L1 ... Training model for up to 0.3s of the 0.3s of remaining time.
    Ran out of time, early stopping on iteration 1. Best iteration is:
    [1]     train_set's multi_error: 0.7289 valid_set's multi_error: 0.816327
    Time limit exceeded... Skipping LightGBMLarge_BAG_L1.
Completed 1/20 k-fold bagging repeats ...
Fitting model: WeightedEnsemble_L2 ... Training model for up to 29.41s of the 0.1s of remaining time.
    0.3558   = Validation accuracy score
    0.33s    = Training runtime
    0.0s     = Validation runtime
AutoGluon training complete, total runtime = 30.24s ...
Fitting 1 L1 models ...
Fitting model: LightGBMXT_BAG_L1_FULL ...
    0.39s    = Training runtime
Deleting model NeuralNetFastAI_BAG_L1. All files under AutogluonModels/ag-20210617_033321/models/NeuralNetFastAI_BAG_L1/ will be removed.
Deleting model LightGBMXT_BAG_L1. All files under AutogluonModels/ag-20210617_033321/models/LightGBMXT_BAG_L1/ will be removed.
Deleting model LightGBM_BAG_L1. All files under AutogluonModels/ag-20210617_033321/models/LightGBM_BAG_L1/ will be removed.
Deleting model RandomForestGini_BAG_L1. All files under AutogluonModels/ag-20210617_033321/models/RandomForestGini_BAG_L1/ will be removed.
Deleting model RandomForestEntr_BAG_L1. All files under AutogluonModels/ag-20210617_033321/models/RandomForestEntr_BAG_L1/ will be removed.
Deleting model ExtraTreesGini_BAG_L1. All files under AutogluonModels/ag-20210617_033321/models/ExtraTreesGini_BAG_L1/ will be removed.
Deleting model ExtraTreesEntr_BAG_L1. All files under AutogluonModels/ag-20210617_033321/models/ExtraTreesEntr_BAG_L1/ will be removed.
Deleting model XGBoost_BAG_L1. All files under AutogluonModels/ag-20210617_033321/models/XGBoost_BAG_L1/ will be removed.
Deleting model NeuralNetMXNet_BAG_L1. All files under AutogluonModels/ag-20210617_033321/models/NeuralNetMXNet_BAG_L1/ will be removed.
Deleting model WeightedEnsemble_L2. All files under AutogluonModels/ag-20210617_033321/models/WeightedEnsemble_L2/ will be removed.
TabularPredictor saved. To load, use: predictor = TabularPredictor.load("AutogluonModels/ag-20210617_033321/")

Another option is to specify more lightweight hyperparameters:

predictor_light = TabularPredictor(label=label, eval_metric=metric).fit(train_data, hyperparameters='very_light', time_limit=30)
No path specified. Models will be saved in: "AutogluonModels/ag-20210617_033351/"
Beginning AutoGluon training ... Time limit = 30s
AutoGluon will save models to "AutogluonModels/ag-20210617_033351/"
AutoGluon Version:  0.2.1b20210617
Train Data Rows:    500
Train Data Columns: 14
Preprocessing data ...
AutoGluon infers your prediction problem is: 'multiclass' (because dtype of label-column == object).
    First 10 (of 15) unique label values:  [' Exec-managerial', ' Other-service', ' Craft-repair', ' Sales', ' Prof-specialty', ' Protective-serv', ' ?', ' Adm-clerical', ' Machine-op-inspct', ' Tech-support']
    If 'multiclass' 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'])
Warning: Some classes in the training set have fewer than 10 examples. AutoGluon will only keep 12 out of 15 classes for training and will not try to predict the rare classes. To keep more classes, increase the number of datapoints from these rare classes in the training data or reduce label_count_threshold.
Fraction of data from classes with at least 10 examples that will be kept for training models: 0.978
Train Data Class Count: 12
Using Feature Generators to preprocess the data ...
Fitting AutoMLPipelineFeatureGenerator...
    Available Memory:                    21687.94 MB
    Train Data (Original)  Memory Usage: 0.28 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', 'relationship', 'race', ...]
    Types of features in processed data (raw dtype, special dtypes):
            ('category', []) : 8 | ['workclass', 'education', 'marital-status', 'relationship', 'race', ...]
            ('int', [])      : 6 | ['age', 'fnlwgt', 'education-num', 'capital-gain', 'capital-loss', ...]
    0.6s = 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 = 0.58s ...
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: 391, Val Rows: 98
Fitting 6 L1 models ...
Fitting model: NeuralNetFastAI ... Training model for up to 29.42s of the 29.42s of remaining time.
No improvement since epoch 6: early stopping
    0.3061   = Validation accuracy score
    1.1s     = Training runtime
    0.02s    = Validation runtime
Fitting model: LightGBM ... Training model for up to 28.29s of the 28.29s of remaining time.
    0.3673   = Validation accuracy score
    0.73s    = Training runtime
    0.01s    = Validation runtime
Fitting model: LightGBMXT ... Training model for up to 27.53s of the 27.53s of remaining time.
    0.3673   = Validation accuracy score
    0.53s    = Training runtime
    0.01s    = Validation runtime
Fitting model: CatBoost ... Training model for up to 26.98s of the 26.97s of remaining time.
    0.3571   = Validation accuracy score
    7.74s    = Training runtime
    0.01s    = Validation runtime
Fitting model: XGBoost ... Training model for up to 19.22s of the 19.22s of remaining time.
    0.398    = Validation accuracy score
    0.72s    = Training runtime
    0.01s    = Validation runtime
Fitting model: NeuralNetMXNet ... Training model for up to 18.39s of the 18.39s of remaining time.
    0.3571   = Validation accuracy score
    5.03s    = Training runtime
    0.02s    = Validation runtime
Fitting model: WeightedEnsemble_L2 ... Training model for up to 29.42s of the 13.06s of remaining time.
    0.4184   = Validation accuracy score
    0.16s    = Training runtime
    0.0s     = Validation runtime
AutoGluon training complete, total runtime = 17.12s ...
TabularPredictor saved. To load, use: predictor = TabularPredictor.load("AutogluonModels/ag-20210617_033351/")

Here you can set hyperparameters to either ‘light’, ‘very_light’, or ‘toy’ to obtain progressively smaller (but less accurate) models and predictors. Advanced users may instead try manually specifying particular models’ hyperparameters in order to make them faster/smaller.

Finally, you may also exclude specific unwieldy models from being trained at all. Below we exclude models that tend to be slower (K Nearest Neighbors, Neural Network, models with custom larger-than-default hyperparameters):

excluded_model_types = ['KNN', 'NN', 'custom']
predictor_light = TabularPredictor(label=label, eval_metric=metric).fit(train_data, excluded_model_types=excluded_model_types, time_limit=30)
No path specified. Models will be saved in: "AutogluonModels/ag-20210617_033409/"
Beginning AutoGluon training ... Time limit = 30s
AutoGluon will save models to "AutogluonModels/ag-20210617_033409/"
AutoGluon Version:  0.2.1b20210617
Train Data Rows:    500
Train Data Columns: 14
Preprocessing data ...
AutoGluon infers your prediction problem is: 'multiclass' (because dtype of label-column == object).
    First 10 (of 15) unique label values:  [' Exec-managerial', ' Other-service', ' Craft-repair', ' Sales', ' Prof-specialty', ' Protective-serv', ' ?', ' Adm-clerical', ' Machine-op-inspct', ' Tech-support']
    If 'multiclass' 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'])
Warning: Some classes in the training set have fewer than 10 examples. AutoGluon will only keep 12 out of 15 classes for training and will not try to predict the rare classes. To keep more classes, increase the number of datapoints from these rare classes in the training data or reduce label_count_threshold.
Fraction of data from classes with at least 10 examples that will be kept for training models: 0.978
Train Data Class Count: 12
Using Feature Generators to preprocess the data ...
Fitting AutoMLPipelineFeatureGenerator...
    Available Memory:                    21668.34 MB
    Train Data (Original)  Memory Usage: 0.28 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', 'relationship', 'race', ...]
    Types of features in processed data (raw dtype, special dtypes):
            ('category', []) : 8 | ['workclass', 'education', 'marital-status', 'relationship', 'race', ...]
            ('int', [])      : 6 | ['age', 'fnlwgt', 'education-num', 'capital-gain', 'capital-loss', ...]
    0.6s = 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 = 0.59s ...
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: 391, Val Rows: 98
Excluded Model Types: ['KNN', 'NN', 'custom']
    Found 'NN' model in hyperparameters, but 'NN' is present in excluded_model_types and will be removed.
    Found 'KNN' model in hyperparameters, but 'KNN' is present in excluded_model_types and will be removed.
    Found 'KNN' model in hyperparameters, but 'KNN' is present in excluded_model_types and will be removed.
Fitting 10 L1 models ...
Fitting model: NeuralNetFastAI ... Training model for up to 29.41s of the 29.41s of remaining time.
    0.3163   = Validation accuracy score
    1.19s    = Training runtime
    0.02s    = Validation runtime
Fitting model: LightGBMXT ... Training model for up to 28.2s of the 28.19s of remaining time.
    0.3673   = Validation accuracy score
    0.55s    = Training runtime
    0.01s    = Validation runtime
Fitting model: LightGBM ... Training model for up to 27.62s of the 27.62s of remaining time.
    0.3673   = Validation accuracy score
    0.72s    = Training runtime
    0.01s    = Validation runtime
Fitting model: RandomForestGini ... Training model for up to 26.88s of the 26.87s of remaining time.
    0.3163   = Validation accuracy score
    0.71s    = Training runtime
    0.11s    = Validation runtime
Fitting model: RandomForestEntr ... Training model for up to 26.03s of the 26.02s of remaining time.
    0.2857   = Validation accuracy score
    0.62s    = Training runtime
    0.11s    = Validation runtime
Fitting model: CatBoost ... Training model for up to 25.28s of the 25.28s of remaining time.
    0.3571   = Validation accuracy score
    7.75s    = Training runtime
    0.01s    = Validation runtime
Fitting model: ExtraTreesGini ... Training model for up to 17.51s of the 17.51s of remaining time.
    0.2857   = Validation accuracy score
    0.72s    = Training runtime
    0.11s    = Validation runtime
Fitting model: ExtraTreesEntr ... Training model for up to 16.66s of the 16.66s of remaining time.
    0.2653   = Validation accuracy score
    0.71s    = Training runtime
    0.11s    = Validation runtime
Fitting model: XGBoost ... Training model for up to 15.81s of the 15.81s of remaining time.
    0.398    = Validation accuracy score
    0.71s    = Training runtime
    0.01s    = Validation runtime
Fitting model: LightGBMLarge ... Training model for up to 14.99s of the 14.99s of remaining time.
    0.3163   = Validation accuracy score
    3.05s    = Training runtime
    0.01s    = Validation runtime
Fitting model: WeightedEnsemble_L2 ... Training model for up to 29.41s of the 11.02s of remaining time.
    0.4184   = Validation accuracy score
    0.26s    = Training runtime
    0.0s     = Validation runtime
AutoGluon training complete, total runtime = 19.26s ...
TabularPredictor saved. To load, use: predictor = TabularPredictor.load("AutogluonModels/ag-20210617_033409/")

If you encounter memory issues

To reduce memory usage during training, you may try each of the following strategies individually or combinations of them (these may harm accuracy):

  • In fit(), set num_bag_sets = 1 (can also try values greater than 1 to harm accuracy less).

  • In fit(), set excluded_model_types = ['KNN', 'XT' ,'RF'] (or some subset of these models).

  • Try different presets in fit().

  • In fit(), set hyperparameters = 'light' or hyperparameters = 'very_light'.

  • Text fields in your table require substantial memory for N-gram featurization. To mitigate this in fit(), you can either: (1) add 'ignore_text' to your presets list (to ignore text features), or (2) specify the argument:

from sklearn.feature_extraction.text import CountVectorizer
from autogluon.features.generators import AutoMLPipelineFeatureGenerator
feature_generator = AutoMLPipelineFeatureGenerator(vectorizer=CountVectorizer(min_df=30, ngram_range=(1, 3), max_features=MAX_NGRAM, dtype=np.uint8))

where MAX_NGRAM = 1000 say (try various values under 10000 to reduce the number of N-gram features used to represent each text field)

In addition to reducing memory usage, many of the above strategies can also be used to reduce training times.

To reduce memory usage during inference:

  • If trying to produce predictions for a large test dataset, break the test data into smaller chunks as demonstrated in FAQ.

  • If models have been previously persisted in memory but inference-speed is not a major concern, call predictor.unpersist_models().

  • If models have been previously persisted in memory, bagging was used in fit(), and inference-speed is a concern: call predictor.refit_full() and use one of the refit-full models for prediction (ensure this is the only model persisted in memory).

If you encounter disk space issues

To reduce disk usage, you may try each of the following strategies individually or combinations of them:

  • Make sure to delete all predictor.path folders from previous fit() runs! These can eat up your free space if you call fit() many times. If you didn’t specify path, AutoGluon still automatically saved its models to a folder called: “AutogluonModels/ag-[TIMESTAMP]”, where TIMESTAMP records when fit() was called, so make sure to also delete these folders if you run low on free space.

  • Call predictor.save_space() to delete auxiliary files produced during fit().

  • Call predictor.delete_models(models_to_keep='best', dry_run=False) if you only intend to use this predictor for inference going forward (will delete files required for non-prediction-related functionality like fit_summary).

  • In fit(), you can add 'optimize_for_deployment' to the presets list, which will automatically invoke the previous two strategies after training.

  • Most of the above strategies to reduce memory usage will also reduce disk usage (but may harm accuracy).

References

The following paper describes how AutoGluon internally operates on tabular data:

Erickson et al. AutoGluon-Tabular: Robust and Accurate AutoML for Structured Data. Arxiv, 2020.