fix #6949

[LittleLittleCloud]

We are excited to review your PR.

So we can do the best job, please check:

• There's a descriptive title that will make sense to other developers some time from now.
• There's associated issues. All PR's should have issue(s) associated - unless a trivial self-evident change such as fixing a typo. You can use the format Fixes #nnnn in your description to cause GitHub to automatically close the issue(s) when your PR is merged.
• Your change description explains what the change does, why you chose your approach, and anything else that reviewers should know.
• You have included any necessary tests in the same PR.

Fix #6949

[ericstj]

Interesting, the addition of SearchSpace sub-namespace now causes a type name resolution problem:

machinelearning/src/Microsoft.ML.TorchSharp/NasBert/Models/NasBertEncoder.cs

Lines 258 to 264 in d0d8569

	.Select(j => SearchSpace.ArchHiddenSize[DiscreteArches[j]]).ToArray();
	var nextHiddenSize = SearchSpace.CheckHiddenDimensionsAndReturnMax(hiddenSizesPerBlock);
	if (nextHiddenSize == 0)
	{
	if (hiddenSizePerBlock.Count == 0)
	{
	nextHiddenSize = SearchSpace.ArchHiddenSize[^1];

Looks like those were meant to refer to this internal class:

machinelearning/src/Microsoft.ML.TorchSharp/NasBert/Modules/SearchSpace.cs

Line 13 in d0d8569

internal static class SearchSpace

Maybe we just rename that @michaelgsharp?

[codecov]

Codecov Report

Attention: 33 lines in your changes are missing coverage. Please review.

Comparison is base (d0d8569) 68.80% compared to head (64a8a73) 68.80%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #6951      +/-   ##
==========================================
- Coverage   68.80%   68.80%   -0.01%     
==========================================
  Files        1249     1249              
  Lines      249644   249686      +42     
  Branches    25481    25485       +4     
==========================================
+ Hits       171778   171797      +19     
- Misses      71275    71294      +19     
- Partials     6591     6595       +4     

Flag	Coverage Δ
Debug	68.80% <67.96%> (-0.01%)	⬇️
production	63.27% <67.96%> (-0.01%)	⬇️
test	88.41% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

Files	Coverage Δ
...oft.ML.Core/SearchSpace/BoolearnChoiceAttribute.cs	100.00% <100.00%> (ø)
...c/Microsoft.ML.Core/SearchSpace/ChoiceAttribute.cs	100.00% <100.00%> (ø)
...crosoft.ML.Core/SearchSpace/NestOptionAttribute.cs	100.00% <100.00%> (ø)
...oft.ML.TorchSharp/NasBert/Models/NasBertEncoder.cs	69.81% <66.66%> (ø)
src/Microsoft.ML.SearchSpace/SearchSpace.cs	82.18% <56.00%> (-2.97%)	⬇️
...rc/Microsoft.ML.Core/SearchSpace/RangeAttribute.cs	54.34% <54.34%> (ø)

... and 1 file with indirect coverage changes

[michaelgsharp]

@ericstj any classes directly involved in the TorchSharp model itself need to match names exactly with the name that was used to save the weights file. We could try and rename this file and see if things still load, but if its at all saved in the weights file its not a class that we can rename. I'll test that out and add a comment here about it.

[michaelgsharp]

Looks good to me.

[michaelgsharp]

/backport to release/3.0

[github-actions]

Started backporting to release/3.0: https://github.com/dotnet/machinelearning/actions/runs/7481061541

[ericstj]

any classes directly involved in the TorchSharp model itself need to match names exactly with the name that was used to save the weights file.

How would we be persisting the name of this static class in our saved file? Still might be worth renaming since we try to avoid name collisions in our codebases.