titanic_xgboost

Titanic survival classification model

This demo utilises a simple model to predict if passenger survived in Titanic disaster given information about his age, sex, passenger's class, ticket fare and number of parent/children abroad the Titanic.

It is trained on data from famous the Kaggle Titanic competition.

Directory structure

• data — Folder contains data to train the model.
• demo — Folder contains a sample Jupyter notebook for invoking a deployed model.
• model — Folder contains model artifacts, ready to be uploaded to the Hydrosphere.
• ops — Folder contains models' training script.

Prerequisites

In order to upload the model to the Hydrosphere you will need the Hydrosphere CLI.

pip install hs

Once you've installed CLI, add your Hydrosphere cluster.

hs cluster add --server http://localhost --name local
hs cluster use local

Model upload

To upload the model, follow below steps.

cd model
hs upload

Model deployment

To deploy a model, create an application from it. You can do it either from the UI, or by using our Python SDK.

from hydrosdk.application import ApplicationBuilder, ExecutionStageBuilder
from hydrosdk import Cluster, ModelVersion
from grpc import ssl_channel_credentials

cluster = Cluster(
    http_address="<hydrosphere-http-address>",
    grpc_address="<hydrosphere-grpc-address>",
    ssl=True,                                       # turn off, if your Hydrosphere instance doesn't have
    grpc_credentials=ssl_channel_credentials(),     # TLS certificates installed
)

mv = ModelVersion.find(cluster, "titanic_xgboost", 1)
mv.lock_till_released()
stage = ExecutionStageBuilder().with_model_variant(mv, 100).build()
app = ApplicationBuilder(cluster, "titanic_xgboost").with_stage(stage).build()
app.lock_while_starting()