Skip to content

sklearn

Directory actions

More options

Directory actions

More options

Latest commit

 

History

History

sklearn

README.md

mlflow-spark-summit-2019 - sklearn

Overview

  • Wine Quality Elastic Net Example
  • This example demonstrates all features of MLflow training and prediction.
  • Saves model in pickle format
  • Saves plot artifacts
  • Shows several ways to run training - mlflow run, run against Databricks cluster, call egg from notebook, etc.
  • Shows several ways to run prediction - web server, mlflow.load_model(), UDF, etc.
  • Data: data/wine-quality-white.csv and data/predict wine-quality-red.csv.

Training

Source: main.py and train.py.

Unmanaged without mlflow run

Command-line python

To run with standard main function:

python main.py --experiment_name sklearn \
  --data_path data/wine-quality-white.csv \
  --alpha 0.5 --l1_ratio 0.5

Jupyter notebook

See Train_Wine_Quality.ipynb.

export MLFLOW_TRACKING_URI=http://localhost:5000
jupyter notebook

Using mlflow run

These runs use the MLproject file. For more details see MLflow documentation - Running Projects.

Note that mlflow run ignores the set_experiment() function so you must specify the experiment with the --experiment-id argument.

mlflow run local

mlflow run . -P alpha=0.01 -P l1_ratio=0.75 -P run_origin=LocalRun --experiment-id=2

mlflow run github

mlflow run https://github.com/amesar/mlflow-fun.git#examples/scikit-learn/wine-quality \
  -P alpha=0.01 -P l1_ratio=0.75 -P run_origin=GitRun \
  --experiment-id=2

mlflow run Databricks remote - Run against Databricks.

See Remote Execution on Databricks and mlflow_run_cluster.json.

Setup.

export MLFLOW_TRACKING_URI=databricks

The token and tracking server URL will be picked up from your Databricks CLI ~/.databrickscfg default profile.

Now run.

mlflow run https://github.com/amesar/mlflow-fun.git#examples/scikit-learn/wine-quality \
  -P alpha=0.01 -P l1_ratio=0.75 -P run_origin=GitRun \
  -P data_path=/dbfs/tmp/data/wine-quality-white.csv \
  --experiment-id=2019 \
  --mode databricks --cluster-spec mlflow_run_cluster.json

Databricks Cluster Runs

You can also package your code as an egg and run it with the standard Databricks REST API endpoints job/runs/submit or jobs/run-now using the spark_python_task.

Setup

Build the egg.

python setup.py bdist_egg

Upload the data file, main file and egg to your Databricks cluster.

databricks fs cp main.py dbfs:/tmp/jobs/wine_quality/main.py
databricks fs cp data/wine-quality-white.csv dbfs:/tmp/jobs/wine_quality/wine-quality-white.csv
databricks fs cp \
  dist/mlflow_wine_quality-0.0.1-py3.6.egg \
  dbfs:/tmp/jobs/wine_quality/mlflow_wine_quality-0.0.1-py3.6.egg

Run Submit

Run with new cluster

Define your run in run_submit_new_cluster.json and launch the run.

curl -X POST -H "Authorization: Bearer MY_TOKEN" \
  -d @run_submit_new_cluster.json  \
  https://myshard.cloud.databricks.com/api/2.0/jobs/runs/submit
Run with existing cluster

Every time you build a new egg, you need to upload (as described above) it to DBFS and restart the cluster.

databricks clusters restart --cluster-id 1222-015510-grams64

Define your run in run_submit_existing_cluster.json and launch the run.

curl -X POST -H "Authorization: Bearer MY_TOKEN" \
  -d @run_submit_existing_cluster.json  \
  https://myshard.cloud.databricks.com/api/2.0/jobs/runs/submit

Job Run Now

Run with new cluster

First create a job with the spec file create_job_new_cluster.json.

databricks jobs create --json-file create_job_new_cluster.json

Then run the job with desired parameters.

databricks jobs run-now --job-id $JOB_ID --python-params ' [ "WineQualityExperiment", 0.3, 0.3, "/dbfs/tmp/jobs/wine_quality/wine-quality-white.csv" ] '
Run with existing cluster

First create a job with the spec file create_job_existing_cluster.json.

databricks jobs create --json-file create_job_existing_cluster.json

Then run the job with desired parameters.

databricks jobs run-now --job-id $JOB_ID --python-params ' [ "WineQualityExperiment", 0.3, 0.3, "/dbfs/tmp/jobs/wine_quality/wine-quality-white.csv" ] '

Run egg from Databricks notebook

Create a notebook with the following cell. Attach it to the existing cluster described above.

from wine_quality import Trainer
data_path = "/dbfs/tmp/jobs/wine_quality/wine-quality-white.csv"
trainer = Trainer("WineQualityExperiment", data_path, "from_notebook_with_egg")
trainer.train(0.4, 0.4)

Predictions

You can make predictions in the following ways:

  1. Use MLflow's serving web server and submit predictions via HTTP calls
  2. Call mlflow.sklearn.load_model() from your own serving code and then make predictions
  3. Call mlflow.pyfunc.load_pyfunc() from your own serving code and then make predictions
  4. Batch prediction with Spark UDF (user-defined function)

See MLflow documentation:

Data for predictions

data/wine-quality-red.csv:

[
  {
    "fixed acidity": 7,
    "volatile acidity": 0.27,
    "citric acid": 0.36,
    "residual sugar": 20.7,
    "chlorides": 0.045,
    "free sulfur dioxide": 45,
    "total sulfur dioxide": 170,
    "density": 1.001,
    "pH": 3,
    "sulphates": 0.45,
    "alcohol": 8.8
  }, 
  . . . . .
]

1. Serving Models from MLflow Web Server

In one window run the server.

mlflow pyfunc serve -p 5001 -r 7e674524514846799310c41f10d6b99d -m model

In another window, submit a prediction.

curl -X POST -H "Content-Type:application/json" -d @data/wine-quality-red.csv http://localhost:5001/invocations

[
    5.551096337521979,
    5.297727513113797,
    5.427572126267637,
    5.562886443251915,
    5.562886443251915
]

2. Predict with mlflow.sklearn.load_model()

python scikit_predict.py 7e674524514846799310c41f10d6b99d

predictions: [5.55109634 5.29772751 5.42757213 5.56288644 5.56288644]

From scikit_predict.py:

model = mlflow.sklearn.load_model("model",run_id="7e674524514846799310c41f10d6b99d")
df = pd.read_csv("data/wine-quality-red.csv")
predicted = model.predict(df)
print("predicted:",predicted)

3. Predict with mlflow.pyfunc.load_pyfunc()

python pyfunc_predict.py 7e674524514846799310c41f10d6b99d

predictions: [5.55109634 5.29772751 5.42757213 5.56288644 5.56288644]

From pyfunc_predict.py:

model_uri = mlflow.start_run("7e674524514846799310c41f10d6b99d").info.artifact_uri +  "/model"
model = mlflow.pyfunc.load_pyfunc(model_uri)
df = pd.read_csv("data/wine-quality-red.csv")
predicted = model.predict(df)
print("predicted:",predicted)

4. Batch prediction with Spark UDF (user-defined function)

Scroll right to see prediction column.

pip install pyarrow

spark-submit --master local[2] spark_udf_predict.py 7e674524514846799310c41f10d6b99d

+-------+---------+-----------+-------+-------------+-------------------+----+--------------+---------+--------------------+----------------+------------------+
|alcohol|chlorides|citric acid|density|fixed acidity|free sulfur dioxide|  pH|residual sugar|sulphates|total sulfur dioxide|volatile acidity|        prediction|
+-------+---------+-----------+-------+-------------+-------------------+----+--------------+---------+--------------------+----------------+------------------+
|    8.8|    0.045|       0.36|  1.001|          7.0|               45.0| 3.0|          20.7|     0.45|               170.0|            0.27| 5.551096337521979|
|    9.5|    0.049|       0.34|  0.994|          6.3|               14.0| 3.3|           1.6|     0.49|               132.0|             0.3| 5.297727513113797|
|   10.1|     0.05|        0.4| 0.9951|          8.1|               30.0|3.26|           6.9|     0.44|                97.0|            0.28| 5.427572126267637|
|    9.9|    0.058|       0.32| 0.9956|          7.2|               47.0|3.19|           8.5|      0.4|               186.0|            0.23| 5.562886443251915|

From spark_udf_predict.py:

spark = SparkSession.builder.appName("ServePredictions").getOrCreate()
df = spark.read.option("inferSchema",True).option("header", True).csv("data/wine-quality-red.csv")
df = df.drop("quality")

udf = mlflow.pyfunc.spark_udf(spark, "model", run_id="7e674524514846799310c41f10d6b99d")
df2 = df.withColumn("prediction", udf(*df.columns))
df2.show(10)