The Orca LibraryΒΆ

Most AI projects start with a Python notebook running on a single laptop; however, one usually needs to go through a mountain of pains to scale it to handle larger data set in a distributed fashion. The Orca library seamlessly scales out your single node Python notebook across large clusters (so as to process distributed Big Data).

First, initialize Orca Context:

from zoo.orca import init_orca_context

# cluster_mode can be "local", "k8s" or "yarn"
sc = init_orca_context(cluster_mode="yarn", cores=4, memory="10g", num_nodes=2) 

Next, perform data-parallel processing in Orca (supporting standard Spark Dataframes, TensorFlow Dataset, PyTorch DataLoader, Pandas, etc.):

from pyspark.sql.functions import array

df =
df = df.withColumn('user', array('user')) \  
       .withColumn('item', array('item'))

Finally, use sklearn-style Estimator APIs in Orca to perform distributed TensorFlow, PyTorch, Keras and BigDL training and inference:

from tensorflow import keras
from import Estimator

user = keras.layers.Input(shape=[1])  
item = keras.layers.Input(shape=[1])  
feat = keras.layers.concatenate([user, item], axis=1)  
predictions = keras.layers.Dense(2, activation='softmax')(feat)  
model = keras.models.Model(inputs=[user, item], outputs=predictions)  

est = Estimator.from_keras(keras_model=model),  
        feature_cols=['user', 'item'],  

See TensorFlow and PyTorch quickstart for more details.