Failed to train model

### URL

https://dotnet.microsoft.com/en-us/learn/ml-dotnet/get-started-tutorial/intro

### Operating system

windows

### More information about your system

Windows 10

### Description

2026-09-10 21:18:54.8852 DEBUG Could not load file or assembly 'Microsoft.CodeAnalysis.CSharp, Version=4.9.0.0, Culture=neutral, PublicKeyToken=31bf3856ad364e35'. The system cannot find the file specified.
at Microsoft.ML.AutoML.Entity.FromExpression(String expression)
at Microsoft.ML.AutoML.Entity.FromExpression(String expression)
at Microsoft.ML.AutoML.PipelineProposer.GetEstimatedCostForPipeline(String schema, SweepablePipeline pipeline)
at Microsoft.ML.AutoML.PipelineProposer.<.ctor>b__12_2(String kv)
at System.Linq.Enumerable.SpanToDictionary[TSource,TKey,TElement](ReadOnlySpan`1 source, Func`2 keySelector, Func`2 elementSelector, IEqualityComparer`1 comparer)
at System.Linq.Enumerable.ToDictionary[TSource,TKey,TElement](IEnumerable`1 source, Func`2 keySelector, Func`2 elementSelector, IEqualityComparer`1 comparer)
at System.Linq.Enumerable.ToDictionary[TSource,TKey,TElement](IEnumerable`1 source, Func`2 keySelector, Func`2 elementSelector)
at Microsoft.ML.AutoML.PipelineProposer..ctor(SweepablePipeline sweepablePipeline, AutoMLExperimentSettings settings)
at Microsoft.ML.AutoML.EciCostFrugalTuner..ctor(SweepablePipeline sweepablePipeline, AutoMLExperimentSettings settings, ITrialResultManager trialResultManager)
at System.Reflection.MethodBaseInvoker.InterpretedInvoke_Constructor(Object obj, IntPtr* args)
at System.Reflection.MethodBaseInvoker.InvokeDirectByRefWithFewArgs(Object obj, Span`1 copyOfArgs, BindingFlags invokeAttr)
at System.Reflection.MethodBaseInvoker.InvokeWithFewArgs(Object obj, BindingFlags invokeAttr, Binder binder, Object[] parameters, CultureInfo culture)
at System.Reflection.RuntimeConstructorInfo.Invoke(BindingFlags invokeAttr, Binder binder, Object[] parameters, CultureInfo culture)
at Microsoft.Extensions.DependencyInjection.ServiceLookup.CallSiteRuntimeResolver.VisitConstructor(ConstructorCallSite constructorCallSite, RuntimeResolverContext context)
at Microsoft.Extensions.DependencyInjection.ServiceLookup.CallSiteRuntimeResolver.VisitRootCache(ServiceCallSite callSite, RuntimeResolverContext context)
at Microsoft.Extensions.DependencyInjection.ServiceLookup.CallSiteVisitor`2.VisitCallSite(ServiceCallSite callSite, TArgument argument)
at Microsoft.Extensions.DependencyInjection.ServiceLookup.CallSiteRuntimeResolver.Resolve(ServiceCallSite callSite, ServiceProviderEngineScope scope)
at Microsoft.Extensions.DependencyInjection.ServiceProvider.CreateServiceAccessor(ServiceIdentifier serviceIdentifier)
at System.Collections.Concurrent.ConcurrentDictionary`2.GetOrAdd(TKey key, Func`2 valueFactory)
at Microsoft.Extensions.DependencyInjection.ServiceProvider.GetService(ServiceIdentifier serviceIdentifier, ServiceProviderEngineScope serviceProviderEngineScope)
at Microsoft.Extensions.DependencyInjection.ServiceProvider.GetService(Type serviceType)
at Microsoft.Extensions.DependencyInjection.ServiceProviderServiceExtensions.GetService[T](IServiceProvider provider)
at Microsoft.ML.AutoML.AutoMLExperiment.RunAsync(CancellationToken ct)
at Microsoft.ML.ModelBuilder.AutoMLService.Experiments.MultiClassificationExperiment.ExecuteAsync(IDataView trainData, IDataView validateData, CancellationToken ct) in /_/src/Microsoft.ML.ModelBuilder.AutoMLService/Experiments/MultiClassificationExperiment.cs:line 140
at Microsoft.ML.ModelBuilder.AutoMLEngine.StartTrainingAsync(ITrainingConfiguration config, PathConfiguration pathConfig, CancellationToken userCancellationToken) in /_/src/Microsoft.ML.ModelBuilder.AutoMLService/AutoMLEngineService/AutoMLEngine.cs:line 178 (Microsoft.ML.ModelBuilder.Utils.Logger.Debug)

### dotnet --info

_No response_

Reproduce the get-started tutorial on Windows 10 and begin with the missing Microsoft.CodeAnalysis.CSharp assembly error. Read AutoMLEngine.cs at line 178 and MultiClassificationExperiment.cs at line 140, then trace the AutoML experiment startup. Done means the tutorial trains successfully without the assembly-load failure.

Written by the indexing model from the issue text.