OPERATOR / VERSION 1.0

piflow_multilayer_perceptron_training

PiFlow · Machine learning

Train a Spark binary multilayer perceptron model on Arrow records and emit the trained model as a ZIP byte stream.

View manifest JSON ↗

Use this operator

operator: piflow_multilayer_perceptron_training@1.0

Configure remote_operator_library_urls with this repository’s origin. XStream verifies and caches artifacts by digest, sharing them across operators.

Named ports

  • sourceInputrecords

    Arrow records with equally sized numeric lists in features. Training also requires numeric class labels in label.

    Formats: Any compatible format
  • resultOutputbytes

    Native Spark ML model packaged as ZIP bytes. Connect to the matching prediction operator.

Output structure and samples are provided by operator inference.

Parameters

  • hiddenLayerSizeintegerRequired

    Width of the single hidden layer from 1 to 256.

    How to fill
    Width of the single hidden layer from 1 to 256. · Minimum: 1 · Maximum: 256
    Example value
    4
  • maxIterintegerRequired

    Maximum optimizer iterations from 1 to 100.

    How to fill
    Maximum optimizer iterations from 1 to 100. · Minimum: 1 · Maximum: 100
    Example value
    8

Operator usage example

Connect the input ports to upstream steps and adjust the parameters for your data.

id: process
operator: piflow_multilayer_perceptron_training@1.0
inputs:
  source: read
params:
  hiddenLayerSize: 4
  maxIter: 8

Manual download

curl -fLo 'operator.jar' 'https://repo.conox.cn/v1/blobs/sha256/e99c6905f4660aadec54ec291743747c1772610e03d346b1d7436ac0d2b60a75/operator.jar'
sha256sum -- 'operator.jar'
curl -fLo 'LICENSE.txt' 'https://repo.conox.cn/v1/blobs/sha256/c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4/LICENSE.txt'
sha256sum -- 'LICENSE.txt'

Compare checksums with the manifest. Download execution and build environments from the runtime catalog ↗.

Source and licensing

See Source archive under Artifact details

Source repository
https://github.com/cas-bigdatalab/piflow
Source revision
777361c4e72b2c5457479683916b7c8e635289ff
License (publisher declared)
Apache-2.0
License text
Available below
Dependency SBOM
Dependency SBOM not provided

Build instructions

Compile src/ using piflow-ecoflow-spark@1.0. See job.yaml for a workflow and fixtures/example.jsonl for example records.

Publisher-provided source, licenses and dependency information support your review. Availability and checksums do not certify safety, legal clearance, or that the executable matches the source.