mmlspark.downloader package

Submodules

mmlspark.downloader.ModelDownloader module

class mmlspark.downloader.ModelDownloader.ModelDownloader(sparkSession, localPath, serverURL='https://mmlspark.azureedge.net/datasets/CNTKModels/')[source]

Bases: object

A class for downloading CNTK pretrained models in python. To download all models use the downloadModels function. To browse models from the microsoft server please use remoteModels.

Parameters

  • sparkSession (SparkSession) – A spark session for interfacing between python and java

  • localPath (str) – The folder to save models to

  • serverURL (str) – The location of the model Server, beware this default can change!

downloadByName(name)[source]

Downloads a named model

Parameters

name (str) – The name of the model

downloadModel(model)[source]

Download a model

Parameters

model (object) – The model to be downloaded

Returns

model schema

Return type

object

downloadModels(models=None)[source]

Download models

Parameters

models – The models to be downloaded

Returns

list of models downloaded

Return type

list

localModels()[source]

Downloads models stored locally on the filesystem

remoteModels()[source]

Downloads models stored remotely.

class mmlspark.downloader.ModelDownloader.ModelSchema(name, dataset, modelType, uri, hash, size, inputNode, numLayers, layerNames)[source]

Bases: object

An object that represents a model.

Parameters

  • name (str) – Name of the model

  • dataset (DataFrame) – Dataset it was trained on

  • modelType (str) – Domain that the model operates on

  • uri (str) – The location of the model’s bytes

  • hash (str) – The sha256 hash of the models bytes

  • size (int) – the size of the model in bytes

  • inputNode (int) – the node which represents the input

  • numLayers (int) – the number of layers of the model

  • layerNames (array) – the names of nodes that represent layers in the network

static fromJava(jobj)[source]
toJava(sparkSession)[source]

Module contents

MicrosoftML is a library of Python classes to interface with the Microsoft scala APIs to utilize Apache Spark to create distibuted machine learning models.

MicrosoftML simplifies training and scoring classifiers and regressors, as well as facilitating the creation of models using the CNTK library, images, and text.