FastTree does not work in UWP

[eerhardt]

System information

• OS version/distro: Windows UWP app
• .NET Version (eg., dotnet --info): UWP

Issue

FastTree is currently always attempting to PrintMemoryStats during training. This is an issue on some platforms (specifically UWP) because these APIs throw exceptions.

We should at least have a way to disable these stats when on platforms that don't support it.

• What did you do?

Try to train a FastTree model in a UWP app.

• What happened?

A PlatformNotSupportedException was thrown from the following method:

machinelearning/src/Microsoft.ML.FastTree/FastTree.cs

Lines 458 to 483 in 6b1a0d3

	protected virtual void PrintMemoryStats(IChannel ch)
	{
	Contracts.AssertValue(ch);
	ch.Trace("Training {0}", GetDatasetStatistics(TrainSet));
	if (ValidSet != null)
	ch.Trace("Validation {0}", GetDatasetStatistics(ValidSet));
	if (TestSets != null)
	{
	for (int i = 0; i < TestSets.Length; ++i)
	ch.Trace("ComputeTests[{1}] {0}",
	GetDatasetStatistics(TestSets[i]), i);
	}
	if (AllowGC)
	ch.Trace("GC Total Memory = {0} MB", GC.GetTotalMemory(true) / 1024 / 1024);
	Process currentProcess = Process.GetCurrentProcess();
	ch.Trace("Working Set = {0} MB", currentProcess.WorkingSet64 / 1024 / 1024);
	ch.Trace("Virtual Memory = {0} MB",
	currentProcess.VirtualMemorySize64 / 1024 / 1024);
	ch.Trace("Private Memory = {0} MB",
	currentProcess.PrivateMemorySize64 / 1024 / 1024);
	ch.Trace("Peak Working Set = {0} MB", currentProcess.PeakWorkingSet64 / 1024 / 1024);
	ch.Trace("Peak Virtual Memory = {0} MB",
	currentProcess.PeakVirtualMemorySize64 / 1024 / 1024);
	}

• What did you expect?

I expected FastTree to work inside a UWP app.

Source code / logs

On UWP apps, you can't get certain information about local processes - it throws an exception:

System.PlatformNotSupportedException: Retrieving information about local processes is not supported on this platform.
   at System.Diagnostics.NtProcessInfoHelper.GetProcessInfos(Predicate`1 processIdFilter)
   at System.Diagnostics.ProcessManager.GetProcessInfo(Int32 processId, String machineName)
   at System.Diagnostics.Process.EnsureState(State state)
   at System.Diagnostics.Process.get_WorkingSet64()

[danmoseley]

I believe all of these will throw PlatformNotSupportedException so one option is just to try/catch for that.

            ch.Trace("Working Set = {0} MB", currentProcess.WorkingSet64 / 1024 / 1024);
            ch.Trace("Virtual Memory = {0} MB",
                currentProcess.VirtualMemorySize64 / 1024 / 1024);
            ch.Trace("Private Memory = {0} MB",
                currentProcess.PrivateMemorySize64 / 1024 / 1024);
            ch.Trace("Peak Working Set = {0} MB", currentProcess.PeakWorkingSet64 / 1024 / 1024);
            ch.Trace("Peak Virtual Memory = {0} MB",
                currentProcess.PeakVirtualMemorySize64 / 1024 / 1024);

Incidentally I also see use of Process.Start to launch a URL (?) in FastTree

System.Diagnostics.Process.Start(uri.AbsoluteUri + "content/GamViz/");

If that's URL I'm surprised if it works on .NET Core. In .NET Core, the default value of UseShellExecute is false. That must be true to launch the browser implicitly from a URL. If indeed that's a problem, you should explicitly set UseShellExecute to true for this launch.

Further note that on UWP, if the value is true, it will throw another PlatformNotSupportedException because it doesn't support UseShellExecute=true. You could catch that, potentially.

[danmoseley]

@eerhardt tells me the Process.Start is not an issue:

The Process.Start call appears to be on an internal-only “visualization” command that is only accessible through the ML.NET command line (which isn’t shipping in v1). It is not reachable from public API, so it shouldn’t be a problem.

• Note: it also depends on being able to start up a web server on the local machine. ☹ Good thing this code is internal.

[Jenscaasen]

Any update on this, or even an ETA?

[FranklinWhale]

I think this issue affects FastTree training on WebAssembly too.

[FranklinWhale]

blazor.webassembly.js:1 
  crit: Microsoft.AspNetCore.Components.WebAssembly.Rendering.WebAssemblyRenderer[100]
      Unhandled exception rendering component: System.Diagnostics.Process is not supported on this platform.
System.PlatformNotSupportedException: System.Diagnostics.Process is not supported on this platform.
   at System.Diagnostics.Process.GetCurrentProcess()
   at Microsoft.ML.Trainers.FastTree.FastTreeTrainerBase`3[[Microsoft.ML.Trainers.FastTree.FastForestBinaryTrainer.Options, Microsoft.ML.FastTree, Version=1.0.0.0, Culture=neutral, PublicKeyToken=cc7b13ffcd2ddd51],[Microsoft.ML.Data.BinaryPredictionTransformer`1[[Microsoft.ML.Trainers.FastTree.FastForestBinaryModelParameters, Microsoft.ML.FastTree, Version=1.0.0.0, Culture=neutral, PublicKeyToken=cc7b13ffcd2ddd51]], Microsoft.ML.Data, Version=1.0.0.0, Culture=neutral, PublicKeyToken=cc7b13ffcd2ddd51],[Microsoft.ML.Trainers.FastTree.FastForestBinaryModelParameters, Microsoft.ML.FastTree, Version=1.0.0.0, Culture=neutral, PublicKeyToken=cc7b13ffcd2ddd51]].PrintMemoryStats(IChannel ch)
   at Microsoft.ML.Trainers.FastTree.FastTreeTrainerBase`3[[Microsoft.ML.Trainers.FastTree.FastForestBinaryTrainer.Options, Microsoft.ML.FastTree, Version=1.0.0.0, Culture=neutral, PublicKeyToken=cc7b13ffcd2ddd51],[Microsoft.ML.Data.BinaryPredictionTransformer`1[[Microsoft.ML.Trainers.FastTree.FastForestBinaryModelParameters, Microsoft.ML.FastTree, Version=1.0.0.0, Culture=neutral, PublicKeyToken=cc7b13ffcd2ddd51]], Microsoft.ML.Data, Version=1.0.0.0, Culture=neutral, PublicKeyToken=cc7b13ffcd2ddd51],[Microsoft.ML.Trainers.FastTree.FastForestBinaryModelParameters, Microsoft.ML.FastTree, Version=1.0.0.0, Culture=neutral, PublicKeyToken=cc7b13ffcd2ddd51]].TrainCore(IChannel ch)
   at Microsoft.ML.Trainers.FastTree.FastForestBinaryTrainer.TrainModelCore(TrainContext context)
   at Microsoft.ML.Trainers.TrainerEstimatorBase`2[[Microsoft.ML.Data.BinaryPredictionTransformer`1[[Microsoft.ML.Trainers.FastTree.FastForestBinaryModelParameters, Microsoft.ML.FastTree, Version=1.0.0.0, Culture=neutral, PublicKeyToken=cc7b13ffcd2ddd51]], Microsoft.ML.Data, Version=1.0.0.0, Culture=neutral, PublicKeyToken=cc7b13ffcd2ddd51],[Microsoft.ML.Trainers.FastTree.FastForestBinaryModelParameters, Microsoft.ML.FastTree, Version=1.0.0.0, Culture=neutral, PublicKeyToken=cc7b13ffcd2ddd51]].TrainTransformer(IDataView trainSet, IDataView validationSet, IPredictor initPredictor)
   at Microsoft.ML.Trainers.TrainerEstimatorBase`2[[Microsoft.ML.Data.BinaryPredictionTransformer`1[[Microsoft.ML.Trainers.FastTree.FastForestBinaryModelParameters, Microsoft.ML.FastTree, Version=1.0.0.0, Culture=neutral, PublicKeyToken=cc7b13ffcd2ddd51]], Microsoft.ML.Data, Version=1.0.0.0, Culture=neutral, PublicKeyToken=cc7b13ffcd2ddd51],[Microsoft.ML.Trainers.FastTree.FastForestBinaryModelParameters, Microsoft.ML.FastTree, Version=1.0.0.0, Culture=neutral, PublicKeyToken=cc7b13ffcd2ddd51]].Fit(IDataView input)
   at BlazorWasmMl.Pages.Index.onLoad() in D:\Documents\GitHub\BlazorWasmMl\Pages\Index.razor:line 40
   at BlazorWasmMl.Pages.Index.OnInitializedAsync() in D:\Documents\GitHub\BlazorWasmMl\Pages\Index.razor:line 153
   at Microsoft.AspNetCore.Components.ComponentBase.RunInitAndSetParametersAsync()

[FranklinWhale]

FastTree training on WebAssembly now works on ML.NET daily build with new FastTreeBinaryTrainer.Options { MemoryStatistics = false }. However, as multithreading is utilized during training, multithreading support on WebAssembly is required, which is only available since .NET 7 RC2.