BrainScript Train, Test, Eval - eldakms/CNTK GitHub Wiki

Here we describe the main top-level commands train and test/eval and their corresponding parameters. For additional top-level commands see the corresponding wiki page.

Train Command

This command asks CNTK to train a model. An example for a train command section is provided on the Config file overview page. The related parameters are:

• reader – the reader configuration block used to determine how to load input data. For details see Reader block.

• SGD – the SGD training setup. For details see SGD block.

• BrainScriptNetworkBuilder – the BrainScript configuration block. For details see BrainScript Network Builder.

• SimpleNetworkBuilder – the simple network builder configuration block. For details see Simple Network Builder.

• cvReader – (optional) the reader configuration block for cross-validation data.

• makeMode – if set to true (default) the training will continue from whatever epoch interrupted. If set to false the training will restart from scratch.

• firstMBsToShowResult – indicates for how many minibatches at the beginning of an epoch to show intermediate results for individually.

• numMBsToShowResult – indicates after how many minibatches the intermediate results should be shown.

Test or Eval Command

These commands evaluate/test a model for accuracy, usually with a test dataset. The related parameters are:

• reader – the reader configuration block to read the test data. For details see Reader block.

• modelPath – the path to the model to be evaluated.

• BrainScriptNetworkBuilder – if this is given, the model is not just read from modelPath but instead constructed from this configuration. This is used to modify models on the fly for evaluation.

• minibatchSize – the minibatch size to use when reading and processing the dataset.

• epochSize – the size of dataset. Default is 0. The entire dataset will be evaluated if it’s set to 0.

• numMBsToShowResult – indicates after how many minibatches the intermediate results should be shown.

• evalNodeNames – an array of one or more node names to evaluate.

• distributedMBReading - accepts Boolean value: true or false; default is false. It should be set to true for readers that support distributed minibatch reading. If you are using CNTK Text Format Reader, Image Reader, or Composite Data Reader, you should set distributedMBReading=true.

The following example is taken form the Simple2d example. In that example the modelPath is defined on the top level and picked up automatically by both the train and test command.

Simple_Demo_Test = [
    action = "test"

    # Parameter values for the reader
    reader = [
        readerType = "CNTKTextFormatReader"
        file = "$DataDir$/SimpleDataTest_cntk_text.txt"
        randomize = false
        input = [
            features = [
                dim = 2        # two-dimensional input data
                format = "dense"
            ]
            labels = [
                dim = 2        # two-dimensional labels
                format = "dense"
            ]
        ]
    ]
]