.net - 如何在 ML.NET 中加载保存的模型？

Question

我正在使用 ML.NET v0.11.0 创建和保存模型。保存过程似乎有效，但我很难确定如何重新加载该模型以供使用。

MLContext mlContext = new MLContext();
IDataView trainingDataView = mlContext.Data.LoadFromEnumerable(myData);

// Setup and train the model
var pipeline = mlContext.Transforms.Conversion.MapValueToKey("Label")
        .Append(mlContext.Transforms.Concatenate("Features", "Field1", "Field2"))
        .AppendCacheCheckpoint(mlContext)
        .Append(mlContext.MulticlassClassification.Trainers.StochasticDualCoordinateAscent(labelColumnName: "Label", featureColumnName: "Features"))
        .Append(mlContext.Transforms.Conversion.MapKeyToValue("PredictedLabel"));

var m = pipeline.Fit(trainingDataView);
using (var stream = System.IO.File.Create("mymodel.zip"))
{
    // this works
    mlContext.Model.Save(m, stream);
}

TransformerChain<KeyToValueMappingTransformer> loadedModel;
using (var stream = System.IO.File.OpenRead("mymodel.zip"))
{
    // this will not work, as ITransformer cannot be cast to TransformerChain<KeyToValueMappingTransformer>
    loadedModel = (TransformerChain<KeyToValueMappingTransformer>)mlContext.Model.Load(stream);
}

Answer 1

我不认为你需要施放负载。试着把它取下来，看看它是否有效。以下对我有用，没有问题。

ITransformer _model;

using (var stream = System.IO.File.OpenRead(modelPath))
{
    _model = _context.Model.Load(stream);
}