ML.NET 发布0.11版本：.NET中的机器学习，为TensorFlow和ONNX添加了新功能

微软发布了其最新版本的机器学习框架：ML.NET0.11带来了新功能和突破性变化。

新版本的机器学习开源框架为TensorFlow和ONNX添加了新功能，但也包括一些重大变化,这也是发布RC版本之前的最后一个预览版，这个月底将发布0.12版本，也就是RC1。

ML.NET的创新0.11

0.11版本的ML.NET现在还支持TensorFlowTransformer组件中的文本输入数据。TensorFlow模型不仅可用于图像，还可用于文本分析。这在.NET博客的代码示例中进行了说明，该博客使用TensorFlow模型进行情感分析：

publicclassTensorFlowSentiment

{

publicstringSentiment_Text;

[VectorType(600)]

publicint[]Features;

[VectorType(2)]

publicfloat[]Prediction;

}

[TensorFlowFact]

publicvoidTensorFlowSentimentClassificationTest()

{

varmlContext=newMLContext(seed:1,conc:1);

vardata=new[]{newTensorFlowSentiment(){Sentiment_Text="thisfilmwasjustbrilliantcastinglocationscenerystorydirectioneveryone'sreallysuitedtheparttheyplayedandyoucouldjustimaginebeingthererobert isanamazingactorandnowthesamebeingdirector fathercamefromthesamescottishislandasmyselfsoilovedthefacttherewasarealconnectionwiththisfilmthewittyremarksthroughoutthefilmweregreatitwasjustbrilliantsomuchthatiboughtthefilmassoonasitwasreleasedfor andwouldrecommendittoeveryonetowatchandtheflyfishingwasamazingreallycriedattheenditwassosadandyouknowwhattheysayifyoucryatafilmitmusthavebeengoodandthisdefinitelywasalso tothetwolittleboy'sthatplayedthe ofnormanandpaultheywerejustbrilliantchildrenareoftenleftoutofthe listithinkbecausethestarsthatplaythemallgrownuparesuchabigprofileforthewholefilmbutthesechildrenareamazingandshouldbepraisedforwhattheyhavedonedon'tyouthinkthewholestorywassolovelybecauseitwastrueandwassomeone'slifeafterallthatwassharedwithusall"}};

vardataView=mlContext.Data.ReadFromEnumerable(data);

varlookupMap=mlContext.Data.ReadFromTextFile(@"sentiment_model/imdb_word_index.csv",

columns:new[]

{

newTextLoader.Column("Words",DataKind.TX,0),

newTextLoader.Column("Ids",DataKind.I4,1),

},

separatorChar:','

);

varestimator=mlContext.Transforms.Text.TokenizeWords("TokenizedWords","Sentiment_Text")

.Append(mlContext.Transforms.Conversion.ValueMap(lookupMap,"Words","Ids",new[]{("Features","TokenizedWords")}));

vardataPipe=estimator.Fit(dataView)

.CreatePredictionEngine<TensorFlowSentiment,TensorFlowSentiment>(mlContext);

stringmodelLocation=@"sentiment_model";

vartfEnginePipe=mlContext.Transforms.ScoreTensorFlowModel(modelLocation,new[]{"Prediction/Softmax"},new[]{"Features"})

.Append(mlContext.Transforms.CopyColumns(("Prediction","Prediction/Softmax")))

.Fit(dataView)

.CreatePredictionEngine<TensorFlowSentiment,TensorFlowSentiment>(mlContext);

//Predictthesentimentforthesampledata

varprocessedData=dataPipe.Predict(data[0]);

Array.Resize(refprocessedData.Features,600);

varprediction=tfEnginePipe.Predict(processedData);

}