Use RAPIDS on a single GPU

This notebook describes a machine learning training workflow using the famous NYC Taxi Dataset. That dataset contains information on taxi trips in New York City.

In this exercise, we attempt to answer this classification question:

based on characteristics that can be known at the beginning of a trip, will this trip result in a high tip?

RAPIDS is a collection of libraries which enable you to take advantage of NVIDIA GPUs to accelerate machine learning workflows. This exercise uses the following RAPIDS packages to execute code on a GPU, rather than a CPU:

• cudf: data frame manipulation, similar to pandas
• cuml: machine learning training and evaluation, similar to scikit-learn

Load data

The code below loads the data into a cudf data frame. This is similar to a pandas dataframe, but it lives in GPU memory and most operations on it are done on the GPU.

import cudf

taxi = cudf.read_csv(
    "https://s3.amazonaws.com/nyc-tlc/trip+data/yellow_tripdata_2019-01.csv",
    parse_dates=["tpep_pickup_datetime", "tpep_dropoff_datetime"],
)

Many dataframe operations that you would execute on a pandas dataframe also work for a cudf dataframe:

len(taxi)

taxi.head()

Train model

Now that the data have been prepped, it’s time to build a model!

For this task, we’ll use the RandomForestClassifier from cuml. If you’ve never used a random forest or need a refresher, consult “Forests of randomized trees” in the scikit-learn documentation. We cast to 32-bit types for compatibility with older versions of cuml.

X = taxi[["PULocationID", "DOLocationID", "passenger_count"]].astype("float32").fillna(-1)
y = (taxi["tip_amount"] > 1).astype("int32")

from cuml.ensemble import RandomForestClassifier

rfc = RandomForestClassifier(n_estimators=100)

_ = rfc.fit(X, y)

Calculate metrics

We’ll use another month of taxi data for the test set and calculate the AUC score

taxi_test = cudf.read_csv(
    "https://s3.amazonaws.com/nyc-tlc/trip+data/yellow_tripdata_2019-02.csv",
    parse_dates=["tpep_pickup_datetime", "tpep_dropoff_datetime"],
)

X_test = taxi_test[["PULocationID", "DOLocationID", "passenger_count"]].astype("float32").fillna(-1)
y_test = (taxi_test["tip_amount"] > 1).astype("int32")

from cuml.metrics import roc_auc_score

preds = rfc.predict_proba(X_test)[1]
roc_auc_score(y_test, preds)