Object reference not set to an instance of an object

[fred-ahead]

System Information (please complete the following information):

• Model Builder or CLI Version: 16.14.1+6737204c7385d9f8d1252a097a5f86747dc0ce81
• Visual Studio Version (if applicable):

Describe the bug

• On which page of the tutorial did you run into an issue (URL): https://dotnet.microsoft.com/en-us/learn/ml-dotnet/get-started-tutorial/train
• Clear description of the problem: When attempting to run the classification command receive the following output: System.NullReferenceException: Object reference not set to an instance of an object. at Microsoft.ML.ModelBuilder.AutoMLService.ServiceFactory.CodeGeneratorService.SetTorchRunTimeFolderAndLoadModel(TrainingConfiguration configuration, String modelPath, MLContext& context, ITransformer& model, DataViewSchema& inputSchema) in //src/Microsoft.ML.ModelBuilder.AutoMLService/ServiceFactory/CodeGeneratorService.cs:line 118 at Microsoft.ML.ModelBuilder.AutoMLService.ServiceFactory.CodeGeneratorService.CreateConsoleAppProjectCodesAsync(TrainingConfiguration configuration, Dictionary`2 sampleData, String sampleProjectName, String nameSpace, String className, String modelPath, String[] labels, String targetFramework, Boolean isDev17, CancellationToken ct) in //src/Microsoft.ML.ModelBuilder.AutoMLService/ServiceFactory/CodeGeneratorService.cs:line 61 at Microsoft.ML.CLI.Program.GenerateConsoleAppAsync(TrainingConfiguration trainingConfiguration, TrainResult trainResult, PathConfiguration pathConfig, String dataset) in //src/mlnet/Program.cs:line 368 at Microsoft.ML.CLI.Program.AutoMLCommandRunner(AutoMLCommand command, Boolean skipGenerateConsoleApp) in //src/mlnet/Program.cs:line 338 at Microsoft.ML.CLI.Program.<>c.<b40>d.MoveNext() in //src/mlnet/Program.cs:line 90 --- End of stack trace from previous location --- at System.CommandLine.Invocation.CommandHandler.GetExitCodeAsync(Object value, InvocationContext context) at System.CommandLine.Invocation.ModelBindingCommandHandler.InvokeAsync(InvocationContext context) at System.CommandLine.Invocation.InvocationPipeline.<>cDisplayClass4_0.<b0>d.MoveNext() --- End of stack trace from previous location --- at System.CommandLine.Builder.CommandLineBuilderExtensions.<>cDisplayClass23_0.<b0>d.MoveNext() --- End of stack trace from previous location --- at Microsoft.ML.CLI.Program.<>cDisplayClass4_0.<b9>d.MoveNext() in /_/src/mlnet/Program.cs:line 291 --- End of stack trace from previous location --- at System.CommandLine.Builder.CommandLineBuilderExtensions.<>c.<b__24_0>d.MoveNext() --- End of stack trace from previous location --- at System.CommandLine.Builder.CommandLineBuilderExtensions.<>cDisplayClass22_0.<b0>d.MoveNext() --- End of stack trace from previous location --- at System.CommandLine.Builder.CommandLineBuilderExtensions.<>cDisplayClass11_0.<b0>d.MoveNext() --- End of stack trace from previous location --- at System.CommandLine.Builder.CommandLineBuilderExtensions.<>c.<b10_0>d.MoveNext() --- End of stack trace from previous location --- at System.CommandLine.Builder.CommandLineBuilderExtensions.<>c__DisplayClass14_0.<b__0>d.MoveNext()

To Reproduce Steps to reproduce the behavior:

1. Run mlnet classification --dataset "yelp_labelled.txt" --label-col 1 --has-header false --name SentimentModel --train-time 60mlnet

Expected behavior A successful build of the classification dataset

[fred-ahead]

Only SentimentModel.mbconfig and SentimentModel.zip is created. The application code is not generated.

[luisquintanilla]

Hi @fred-ahead

Sorry to hear you ran into this issue. We're looking into it.

@LittleLittleCloud can you please take a look at this one. Thanks.

[derekforeman]

I am seeing the same issue mlnet version : 16.14.2+48a74e849ab32700b5af6489a96ff50c0e3bc429

[LittleLittleCloud]

Looking at it

[LittleLittleCloud]

We push a fix for it, in the meanwhile, a workaround can be adding a $Path variable in your system environment.

[fred-ahead]

@LittleLittleCloud What is missing from the path? Happy to do that, just need to know what to add to my path.

[LittleLittleCloud]

@fred-ahead The content of $Path doesn't matter. This issue is caused because mlnet cli can't find $Path environment variable is missing in the system.

[joanimato]

@LittleLittleCloud can you elaborate? I'm running into the same exact issue with the mlnet cli on MacOs. My path variable is set and displays correctly when doing echo $PATH

[LittleLittleCloud]

@joanimato sure. We just push a fix to nuget and please update to the latest cli to deploy the fix.

More details on this error if anyone interests

MLNet cli checks if env:path exists in environment by calling

var path = Environment.GetVariable("Path")

and then adds runtime folder into env:Path.

by calling

if (!path.contain(runtimefolder))
    path = /path/to/runtime;path

The problem is Environment.GetVariable("Path") returns a null if "Path" doesn't exist. And if "Path" is null, calling path.contain will trigger an object reference not set instance exception.

[fred-ahead]

If it checks for Path then that is most likely the issue as on a Mac the PATH variable is all caps. Path does not work for Mac, neither does path. PATH does work though. On Windows the capitalization doesn't matter.

[fred-ahead]

To be clear, on a Mac this issue is easily resolved by doing the following:

export Path=$PATH

I was able to easily build the code with the above command.

[LittleLittleCloud]

it might affect text classification GPU training in Mac. @LittleLittleCloud to look into