-
Notifications
You must be signed in to change notification settings - Fork 1.9k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Change ML.NET to work with .NET Framework 4.6.1 #1075
Conversation
@@ -72,7 +72,7 @@ public EstimatorChain<TNewTrans> Append<TNewTrans>(IEstimator<TNewTrans> estimat | |||
where TNewTrans : class, ITransformer | |||
{ | |||
Contracts.CheckValue(estimator, nameof(estimator)); | |||
return new EstimatorChain<TNewTrans>(_estimators.Append(estimator).ToArray(), _scopes.Append(scope).ToArray()); | |||
return new EstimatorChain<TNewTrans>(_estimators.Concat(new[] { estimator }).ToArray(), _scopes.Concat(new[] { scope }).ToArray()); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'm not sure how concerned we are about perf in these methods, but another way would be to make our own AppendElement
method:
public static T[] AppendElement(T[] array, T element)
{
T[] result = new T[array.Length + 1];
Array.Copy(array, result);
result[array.Length] = element;
return result;
}
Another option would be to use ImmutableCollections.... I don't think you'd have to pay a penalty for copying in that case. But I'm not sure we'd want to take a new dependency, if we aren't using ImmutableCollections already.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Hi @eerhardt -- we are using ImmutableCollections
, and I am beginning to use it in my code. I really like it a lot, I think we ought to use it more.
@@ -342,7 +342,7 @@ public CoefficientStatistics[] GetCoefficientStatistics(LinearBinaryPredictor pa | |||
return null; | |||
|
|||
var order = GetUnorderedCoefficientStatistics(parent, schema).OrderByDescending(stat => stat.ZScore).Take(paramCountCap - 1); | |||
return order.Prepend(new CoefficientStatistics("(Bias)", bias, stdError, zScore, pValue)).ToArray(); | |||
return order.Prepend(new[] { new CoefficientStatistics("(Bias)", bias, stdError, zScore, pValue) }).ToArray(); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Which order.Prepend
method is this?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Ah, we have our own shim
here: src\Microsoft.ML.Core\Data\MetadataUtils.cs
/// <summary>
/// Prepends a params array to an enumerable. Useful when implementing GetMetadataTypes.
/// </summary>
public static IEnumerable<T> Prepend<T>(this IEnumerable<T> tail, params T[] head)
{
return head.Concat(tail);
}
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
We have an overload of this method that takes a params T[], so this still works with older versions of .NET. I thought it would be better to call the same overload independent of the framework, so I changed it.
In reply to: 221026867 [](ancestors = 221026867)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Just a minor comment if we are concerned about perf, which I don't think we are in that method since it shouldn't be called often.
@@ -342,7 +342,7 @@ public CoefficientStatistics[] GetCoefficientStatistics(LinearBinaryPredictor pa | |||
return null; | |||
|
|||
var order = GetUnorderedCoefficientStatistics(parent, schema).OrderByDescending(stat => stat.ZScore).Take(paramCountCap - 1); | |||
return order.Prepend(new CoefficientStatistics("(Bias)", bias, stdError, zScore, pValue)).ToArray(); | |||
return order.Prepend(new[] { new CoefficientStatistics("(Bias)", bias, stdError, zScore, pValue) }).ToArray(); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Ah, we have our own shim
here: src\Microsoft.ML.Core\Data\MetadataUtils.cs
/// <summary>
/// Prepends a params array to an enumerable. Useful when implementing GetMetadataTypes.
/// </summary>
public static IEnumerable<T> Prepend<T>(this IEnumerable<T> tail, params T[] head)
{
return head.Concat(tail);
}
Closing/opening to re-check CI tests |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks @yaeldekel
@@ -75,7 +76,7 @@ private static MethodInfo[] InitValueTupleCreateMethods() | |||
var methods = typeof(ValueTuple).GetMethods() | |||
.Where(m => m.Name == methodName && m.ContainsGenericParameters) | |||
.OrderBy(m => m.GetGenericArguments().Length).Take(7) | |||
.Append(typeof(AnalyzeUtil).GetMethod(nameof(UnstructedCreate))).ToArray(); | |||
.ToArray().AppendElement(typeof(AnalyzeUtil).GetMethod(nameof(UnstructedCreate))); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It feels wrong to allocate an array and immediate throw it away to allocate one 1 element longer.
@@ -146,7 +146,7 @@ public TransformerChain<TNewLast> Append<TNewLast>(TNewLast transformer, Transfo | |||
where TNewLast : class, ITransformer | |||
{ | |||
Contracts.CheckValue(transformer, nameof(transformer)); | |||
return new TransformerChain<TNewLast>(_transformers.Append(transformer).ToArray(), _scopes.Append(scope).ToArray()); | |||
return new TransformerChain<TNewLast>(_transformers.AppendElement(transformer), _scopes.AppendElement(scope)); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Looks like this type only has enumerable constructor, so you end up with 3 arrays for each param. One that is size N, and two that are size N+1. If you keep it as an enumerable, you can reduce this. It can be further reduced if you have a constructor that takes an array with reference semantics.
It's also somewhat odd that TransformerChain has copy semantics for constructor params, but EstimatorChain has reference semantics (and thus doesn't have this array copy problem).
Fixes #1072 .