Automatic Logging with MLflow Tracking

Auto logging is a powerful feature that allows you to log metrics, parameters, and models without the need for explicit log statements. All you need to do is to call mlflow.autolog() before your training code.

import mlflow

mlflow.autolog()

with mlflow.start_run():
    # your training code goes here
    ...

This will enable MLflow to automatically log various information about your run, including:

Metrics - MLflow pre-selects a set of metrics to log, based on what model and library you use
Parameters - hyper params specified for the training, plus default values provided by the library if not explicitly set
Model Signature - logs Model signature instance, which describes input and output schema of the model
Artifacts - e.g. model checkpoints
Dataset - dataset object used for training (if applicable), such as tensorflow.data.Dataset

How to Get started

Step 1 - Get MLflow

MLflow is available on PyPI. If you don’t already have it installed on your system, you can install it with:

pip install mlflow

Step 2 - Insert `mlflow.autolog` in Your Code

For example, following code snippet shows how to enable autologging for a scikit-learn model:

import mlflow

from sklearn.model_selection import train_test_split
from sklearn.datasets import load_diabetes
from sklearn.ensemble import RandomForestRegressor

mlflow.autolog()

db = load_diabetes()
X_train, X_test, y_train, y_test = train_test_split(db.data, db.target)

rf = RandomForestRegressor(n_estimators=100, max_depth=6, max_features=3)
# MLflow triggers logging automatically upon model fitting
rf.fit(X_train, y_train)

Step 3 - Execute Your Code

python YOUR_ML_CODE.py

Step 4 - View Your Results in the MLflow UI

Once your training job finishes, you can run following command to launch the MLflow UI:

mlflow ui --port 8080

Then, navigate to http://localhost:8080 in your browser to view the results.

Customize Autologging Behavior

You can also control the behavior of autologging by passing arguments to mlflow.autolog() function. For example, you can disable logging of model checkpoints and assosiate tags with your run as follows:

import mlflow

mlflow.autolog(
    log_model_signatures=False,
    extra_tags={"YOUR_TAG": "VALUE"},
)

See mlflow.autolog() for the full set of arguments you can use.

Enable / Disable Autologging for Specific Libraries

One common use case is to enable/disable autologging for a specific library. For example, if you train your model on PyTorch but use scikit-learn for data preprocessing, you may want to disable autologging for scikit-learn while keeping it enabled for PyTorch. You can achieve this by either (1) enable autologging only for PyTorch using PyTorch flavor (2) disable autologging for scikit-learn using its flavor with disable=True.

import mlflow

# Option 1: Enable autologging only for PyTorch
mlflow.pytorch.autolog()

# Option 2: Disable autologging for scikit-learn, but enable it for other libraries
mlflow.sklearn.autolog(disable=True)
mlflow.autolog()

Supported Libraries

Note

The generic autolog function mlflow.autolog() enables autologging for each supported library you have installed as soon as you import it. Alternatively, you can use library-specific autolog calls such as mlflow.pytorch.autolog() to explicitly enable (or disable) autologging for a particular library.

The following libraries support autologging:

Fastai
Gluon
Keras/TensorFlow
LangChain
LlamaIndex
LightGBM
OpenAI
Paddle
PySpark
PyTorch
Scikit-learn
Spark
Statsmodels
XGBoost

For flavors that automatically save models as an artifact, additional files for dependency management are logged.

Fastai

Call the generic autolog function mlflow.fastai.autolog() before your training code to enable automatic logging of metrics and parameters. See an example usage with Fastai.

Autologging captures the following information:

Framework	Metrics	Parameters	Tags	Artifacts
fastai	user-specified metrics	Logs optimizer data as parameters. For example, `epochs`, `lr`, `opt_func`, etc; Logs the parameters of the EarlyStoppingCallback and OneCycleScheduler callbacks	–	Model checkpoints are logged to a ‘models’ directory; MLflow Model (fastai Learner model) on training end; Model summary text is logged

Gluon

Call the generic autolog function mlflow.gluon.autolog() before your training code to enable automatic logging of metrics and parameters. See example usages with Gluon .

Autologging captures the following information:

Framework	Metrics	Parameters	Tags	Artifacts
Gluon	Training loss; validation loss; user-specified metrics	Number of layers; optimizer name; learning rate; epsilon	–	MLflow Model (Gluon model); on training end

Keras/TensorFlow

Call the generic autolog function or mlflow.tensorflow.autolog() before your training code to enable automatic logging of metrics and parameters. As an example, try running the Keras/Tensorflow example.

Note that only versions of tensorflow>=2.3 are supported. The respective metrics associated with tf.estimator and EarlyStopping are automatically logged. As an example, try running the Keras/TensorFlow example.

Autologging captures the following information:

Framework/module	Metrics	Parameters	Tags	Artifacts
`tf.keras`	Training loss; validation loss; user-specified metrics	`fit()` parameters; optimizer name; learning rate; epsilon	–	Model summary on training start; MLflow Model (Keras model); TensorBoard logs on training end
`tf.keras.callbacks.EarlyStopping`	Metrics from the `EarlyStopping` callbacks. For example, `stopped_epoch`, `restored_epoch`, `restore_best_weight`, etc	`fit()` parameters from `EarlyStopping`. For example, `min_delta`, `patience`, `baseline`, `restore_best_weights`, etc	–	–

If no active run exists when autolog() captures data, MLflow will automatically create a run to log information to. Also, MLflow will then automatically end the run once training ends via calls to tf.keras.fit().

If a run already exists when autolog() captures data, MLflow will log to that run but not automatically end that run after training. You will have to manually stop the run if you wish to start a new run context for logging to a new run.

LangChain

Call the generic autolog function mlflow.langchain.autolog() before your training code to enable automatic logging of traces. See LangChain Autologging for more details.

Autologging captures the following information:

Framework	Metrics	Parameters	Tags	Artifacts
LangChain	–	–	–	Traces MLflow Model (LangChain model) with model signature on training end Input example

LlamaIndex

Call the generic autolog function mlflow.llama_index.autolog() before your training code to enable automatic logging of traces.

Autologging captures the following information:

Framework	Metrics	Parameters	Tags	Artifacts
LlamaIndex	–	–	–	Traces

LightGBM

Call the generic autolog function mlflow.lightgbm.autolog() before your training code to enable automatic logging of metrics and parameters.

Autologging captures the following information:

Framework	Metrics	Parameters	Tags	Artifacts
LightGBM	user-specified metrics	lightgbm.train parameters	–	MLflow Model (LightGBM model) with model signature on training end; feature importance; input example

If early stopping is activated, metrics at the best iteration will be logged as an extra step/iteration.

OpenAI

Call the generic autolog function mlflow.openai.autolog() before your training code to enable automatic logging of artifacts. See an example usage with OpenAI.

Autologging captures the following information:

Framework	Metrics	Parameters	Tags	Artifacts
OpenAI	–	–	–	MLflow Model (OpenAI model) with model signature on training end Input example

Paddle

Call the generic autolog function mlflow.paddle.autolog() before your training code to enable automatic logging of metrics and parameters.

Autologging captures the following information:

Framework	Metrics	Parameters	Tags	Artifacts
Paddle	user-specified metrics	paddle.Model.fit parameters	–	MLflow Model (Paddle model) with model signature on training end

PySpark

Call mlflow.pyspark.ml.autolog() before your training code to enable automatic logging of metrics, params, and models. See example usage with PySpark.

Autologging for pyspark ml estimators captures the following information:

Metrics	Parameters	Tags	Artifacts
Post training metrics obtained by `Evaluator.evaluate`	Parameters obtained by `Estimator.fit`	Class name Fully qualified class name	MLflow Model containing a fitted estimator `metric_info.json` for post training metrics

PyTorch

Call the generic autolog function mlflow.pytorch.autolog() before your PyTorch Lightning training code to enable automatic logging of metrics, parameters, and models. See example usages here. Note that currently, PyTorch autologging supports only models trained using PyTorch Lightning.

Autologging is triggered on calls to pytorch_lightning.trainer.Trainer.fit and captures the following information:

Framework/module	Metrics	Parameters	Tags	Artifacts
`pytorch_lightning.trainer.Trainer`	Training loss; validation loss; average_test_accuracy; user-defined-metrics.	`fit()` parameters; optimizer name; learning rate; epsilon.	–	Model summary on training start, MLflow Model (PyTorch model) on training end;
`pytorch_lightning.callbacks.earlystopping`	Training loss; validation loss; average_test_accuracy; user-defined-metrics. Metrics from the `EarlyStopping` callbacks. For example, `stopped_epoch`, `restored_epoch`, `restore_best_weight`, etc.	`fit()` parameters; optimizer name; learning rate; epsilon Parameters from the `EarlyStopping` callbacks. For example, `min_delta`, `patience`, `baseline`,``restore_best_weights``, etc	–	Model summary on training start; MLflow Model (PyTorch model) on training end; Best PyTorch model checkpoint, if training stops due to early stopping callback.

If no active run exists when autolog() captures data, MLflow will automatically create a run to log information, ending the run once the call to pytorch_lightning.trainer.Trainer.fit() completes.

If a run already exists when autolog() captures data, MLflow will log to that run but not automatically end that run after training.

Note

Parameters not explicitly passed by users (parameters that use default values) while using pytorch_lightning.trainer.Trainer.fit() are not currently automatically logged
In case of a multi-optimizer scenario (such as usage of autoencoder), only the parameters for the first optimizer are logged

Scikit-learn

Call mlflow.sklearn.autolog() before your training code to enable automatic logging of sklearn metrics, params, and models. See example usage here.

Autologging for estimators (e.g. LinearRegression) and meta estimators (e.g. Pipeline) creates a single run and logs:

Metrics	Parameters	Tags	Artifacts
Training score obtained by `estimator.score`	Parameters obtained by `estimator.get_params`	Class name Fully qualified class name	Fitted estimator

Autologging for parameter search estimators (e.g. GridSearchCV) creates a single parent run and nested child runs

- Parent run
  - Child run 1
  - Child run 2
  - ...

containing the following data:

Run type	Metrics	Parameters	Tags	Artifacts
Parent	Training score	Parameter search estimator’s parameters Best parameter combination	Class name Fully qualified class name	Fitted parameter search estimator Fitted best estimator Search results csv file
Child	CV test score for each parameter combination	Each parameter combination	Class name Fully qualified class name	–

Spark

Initialize a SparkSession with the mlflow-spark JAR attached (e.g. SparkSession.builder.config("spark.jars.packages", "org.mlflow.mlflow-spark")) and then call the generic autolog function mlflow.spark.autolog() to enable automatic logging of Spark datasource information at read-time, without the need for explicit log statements. Note that autologging of Spark ML (MLlib) models is not yet supported.

Autologging captures the following information:

Framework	Metrics	Parameters	Tags	Artifacts
Spark	–	–	Single tag containing source path, version, format. The tag contains one line per datasource	–

Note

Moreover, Spark datasource autologging occurs asynchronously - as such, it’s possible (though unlikely) to see race conditions when launching short-lived MLflow runs that result in datasource information not being logged.

Important

With Pyspark 3.2.0 or above, Spark datasource autologging requires PYSPARK_PIN_THREAD environment variable to be set to false.

Statsmodels

Call the generic autolog function mlflow.statsmodels.autolog() before your training code to enable automatic logging of metrics and parameters.

Autologging captures the following information:

Framework	Metrics	Parameters	Tags	Artifacts
Statsmodels	user-specified metrics	statsmodels.base.model.Model.fit parameters	–	MLflow Model (statsmodels.base.wrapper.ResultsWrapper) on training end

Note

Each model subclass that overrides fit expects and logs its own parameters.

XGBoost

Call the generic autolog function mlflow.xgboost.autolog() before your training code to enable automatic logging of metrics and parameters.

Autologging captures the following information:

Framework	Metrics	Parameters	Tags	Artifacts
XGBoost	user-specified metrics	xgboost.train parameters	–	MLflow Model (XGBoost model) with model signature on training end; feature importance; input example

If early stopping is activated, metrics at the best iteration will be logged as an extra step/iteration.

Automatic Logging with MLflow Tracking

How to Get started

Step 1 - Get MLflow

Step 2 - Insert mlflow.autolog in Your Code

Step 3 - Execute Your Code

Step 4 - View Your Results in the MLflow UI

Customize Autologging Behavior

Enable / Disable Autologging for Specific Libraries

Supported Libraries

Step 2 - Insert `mlflow.autolog` in Your Code