[Create ML Components] The transformer is not representable as a CoreML model (ImageReader).

Question

Following this guide https://developer.apple.com/documentation/CreateML/creating-a-multi-label-image-classifier

Has anyone been able to export a CoreML model, specifically according to the documentation below? Since there isn't any runnable examples, just snippets, perhaps documentation error? If anyone is already familiar with these pipeline generics, is there something that jumps out about the example transformer used that fails conformance or just factually incorrect?

Export the model to use with Vision

After you train the model, you can export it as a Core ML model.

// Export to Core ML
let modelURL = URL(filePath: "/path/to/model")
try model.export(to: modelURL)

Create ML

529

Posted by

rromanchuk

Reply

Add a Comment

Answer 1

Somehow got it to export by composing the transformer explicitly, which gave a ComposedTransformer<ImageFeaturePrint, FullyConnectedNetworkMultiLabelClassifierModel>, so I'm guessing the chaining used in the documentation was never attempted to export since return type is invalid

let annotatedFeatures = detectionFiles.map {
    AnnotatedFeature(
        feature: directoryURL.appending(component: $0.filename),
        annotation: $0.labels
    )
}

let reader = ImageReader()
let (training, validation) = annotatedFeatures.randomSplit(by: 0.8)

let featurePrint = ImageFeaturePrint(revision: 2)
let classifier = FullyConnectedNetworkMultiLabelClassifier<Float, String>(labels: labels)
let task = featurePrint.appending(classifier)

Task {
    let trainingImages = try await reader.applied(to: training)
    let validationImages = try await reader.applied(to: validation)
    
    let model = try await task.fitted(
        to: trainingImages,
        validateOn: validationImages,
        eventHandler: { event in
            debugPrint(event)
        }
    )
    try! model.export(to: modelFile)
}

Posted by

rromanchuk

Add a Comment

Answer 2

Thanks for pointing this out!

Posted by

Frameworks Engineer

Add a Comment

[Create ML Components] The transformer is not representable as a CoreML model (ImageReader).

Replies