CSV folder example

Example 1: directory structure

• bucket_root
  • nested_folder/
    • 2021-05-03/ -> Date Partition
      • training_data/ -> Sub Folder
        • data.csv
    • 2021-05-04/ -> Date Partition
      • training_sample/ -> Sub Folder
        • data.csv

Python

from featurestore import Client, CSVFolder

# Initialise feature store client
client = Client("ip:port")
client.auth.login()

# Set project specifics
project = client.projects.create("demo")

# Create the csv folder source
csv_folder = CSVFolder(
  root_folder="s3a://feature-store-test-data/nested_folder",
  filter_pattern=".*/training.*"
)
csv_folder_schema = client.extract_schema_from_source(csv_folder)

# Register the feature set
my_feature_set = project.feature_sets.register(csv_folder_schema, "feature_set_name", primary_key=["key_name"])

# Ingest to cache
my_feature_set.ingest(csv_folder)

# Retrieve feature set
ref = my_feature_set.retrieve()
ref.download()

Scala

import ai.h2o.featurestore.Client
import ai.h2o.featurestore.core.sources.CSVFolder

// Initialise feature store client
val client = Client("url")
client.auth.login()

// Set project specifics
val project = client.projects.create("demo")

// Create the csv folder source
val csvFolder = CSVFolder(
 rootFolder="s3a://feature-store-test-data/nested_folder",
 filterPattern=".*/training.*"
)
val csvFolderSchema = client.extractSchemaFromSource(csvFolder)

// Register the feature set
val myFeatureSet = project.featureSets.register(csvFolderSchema, "featureSetName", primaryKey=Seq("keyName"))

// Ingest to cache
myFeatureSet.ingest(csvFolder)

# Retrieve feature set
val ref = myFeatureSet.retrieve()
ref.download()

Example 2: directory structure

• bucket_root

  • nested_folder/

    • California

      • 2021-05-03/ -> Date Partition
        • training_data/ -> Sub Folder
          • date.csv

    • Arizona

      • 2021-05-04/ -> Date Partition
        • training_sample/ -> Sub Folder
          • data.csv

    • Texas

      • 2021-05-04/ -> Date Partition
        • training_sample/ -> Sub Folder
          • data.csv

Python

from featurestore import Client, CSVFolder

# Initialise feature store client
client = Client("ip:port")
client.auth.login()

# Set project specifics
project = client.projects.create("demo")

# Create the csv folder source
csv_folder_source = CSVFolder(
  root_folder="s3a://feature-store-test-data/nested_folder",
  filter_pattern=".*/.*/training.*"  # To ingest from all states
)
csv_folder_schema = client.extract_schema_from_source(csv_folder_source)
# Note
# To ingest only from California, then filter_pattern = "California/.*/training.*"
# To ingest only from California & Arizona, then filter_pattern = "(Arizona|California)/.*/training.*"

# Register the feature set
my_feature_set = project.feature_sets.register(csv_folder_schema, "feature_set_name", primary_key=["key_name"])

# Ingest to cache
my_feature_set.ingest()

# Retrieve feature set
ref = my_feature_set.retrieve()
ref.download()

Scala

import ai.h2o.featurestore.Client
import ai.h2o.featurestore.core.sources.CSVFolder

// Initialise feature store client
val client = Client("url")
client.auth.setAuthToken(...)

// Set project specifics
val project = client.projects.create("demo")


// Create the csv folder source
val csvFolderSource = CSVFolder(
  rootFolder="s3a://feature-store-test-data/nested_folder",
  filterPattern=".*/.*/training.*"
)
val csvFolderSchema = client.extractSchemaFromSource(csvFolderSource)
// Note
// To ingest only from California, then filter_pattern = "California/.*/training.*"
// To ingest only from California & Arizona, then filter_pattern = "(Arizona|California)/.*/training.*"

// Register the feature set
val myFeatureSet = project.featureSets.register(csvFolderSchema, "featureSetName", primaryKey=Seq("keyName"))

// Ingest to cache
myFeatureSet.ingest()

// Retrieve feature set
val ref = myFeatureSet.retrieve()
ref.download()

Example 3: directory structure (no date folder)

• bucket_root

  • nested_folder/

    • California

      • training_data/ -> Sub Folder
        • data.csv

    • Arizona

      • training_sample/ -> Sub Folder
        • data.csv

    • Texas

      • training_sample/ -> Sub Folder
        • data.csv

Python

from featurestore import Client, CSVFolder

# Initialise feature store client
client = Client("ip:port")
client.auth.login()

# Set project specifics
project = client.projects.create("demo")

# Create the csv folder source
csv_folder_source = CSVFolder(
  root_folder="s3a://feature-store-test-data/nested_folder",
  filter_pattern=".*/training.*"  # To ingest from all states
)
csv_folder_schema = client.extract_schema_from_source(csv_folder_source)
# Note
# To ingest only from California, then filter_pattern = "California/training.*"
# To ingest only from California & Arizona, then filter_pattern = "(Arizona|California)/training.*"

# Register the feature set
my_feature_set = project.feature_sets.register(csv_folder_schema, "feature_set_name", primary_key=["key_name"])

# Ingest to cache
my_feature_set.ingest()

# Retrieve feature set
ref = my_feature_set.retrieve()
ref.download()

Scala

import ai.h2o.featurestore.Client
import ai.h2o.featurestore.core.sources.CSVFolder

// Initialise feature store client
val client = Client("url")
client.auth.setAuthToken(...)

// Set project specifics
val project = client.projects.create("demo")

// Create the csv folder source
val csvFolderSource = CSVFolder(
  rootFolder="s3a://feature-store-test-data/nested_folder",
  filterPattern=".*/training.*"  # To ingest from all states
)
val csvFolderSchema = client.extractSchemaFromSource(csvFolderSource)
// Note
// To ingest only from California, then filter_pattern = "California/training.*"
// To ingest only from California & Arizona, then filter_pattern = "(Arizona|California)/training.*"

// Register the feature set
val myFeatureSet = project.featureSets.register(csvFolderSchema, "featureSetName", primaryKey=Seq("keyName"))

// Ingest to cache
myFeatureSet.ingest()

// Retrieve feature set
val ref = myFeatureSet.retrieve()
ref.download()

---

Feedback

• Submit and view feedback for this page
• Send feedback about H2O Feature Store to cloud-feedback@h2o.ai