-
Notifications
You must be signed in to change notification settings - Fork 1.9k
/
GcnTransform.cs
1102 lines (986 loc) · 55.3 KB
/
GcnTransform.cs
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
// Licensed to the .NET Foundation under one or more agreements.
// The .NET Foundation licenses this file to you under the MIT license.
// See the LICENSE file in the project root for more information.
using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using Microsoft.ML;
using Microsoft.ML.CommandLine;
using Microsoft.ML.Data;
using Microsoft.ML.EntryPoints;
using Microsoft.ML.Internal.CpuMath;
using Microsoft.ML.Internal.Utilities;
using Microsoft.ML.Model.OnnxConverter;
using Microsoft.ML.Runtime;
using Microsoft.ML.Transforms;
[assembly: LoadableClass(LpNormNormalizingTransformer.GcnSummary, typeof(IDataTransform), typeof(LpNormNormalizingTransformer), typeof(LpNormNormalizingTransformer.GcnOptions), typeof(SignatureDataTransform),
LpNormNormalizingTransformer.UserNameGn, "GcnTransform", LpNormNormalizingTransformer.ShortNameGn)]
[assembly: LoadableClass(LpNormNormalizingTransformer.GcnSummary, typeof(IDataTransform), typeof(LpNormNormalizingTransformer), null, typeof(SignatureLoadDataTransform),
LpNormNormalizingTransformer.UserNameGn, LpNormNormalizingTransformer.LoaderSignature, LpNormNormalizingTransformer.LoaderSignatureOld)]
[assembly: LoadableClass(LpNormNormalizingTransformer.Summary, typeof(IDataTransform), typeof(LpNormNormalizingTransformer), typeof(LpNormNormalizingTransformer.Options), typeof(SignatureDataTransform),
LpNormNormalizingTransformer.UserNameLP, "LpNormNormalizer", LpNormNormalizingTransformer.ShortNameLP)]
[assembly: LoadableClass(LpNormNormalizingTransformer.Summary, typeof(LpNormNormalizingTransformer), null, typeof(SignatureLoadModel),
LpNormNormalizingTransformer.UserNameGn, LpNormNormalizingTransformer.LoaderSignature)]
[assembly: LoadableClass(typeof(IRowMapper), typeof(LpNormNormalizingTransformer), null, typeof(SignatureLoadRowMapper),
LpNormNormalizingTransformer.UserNameGn, LpNormNormalizingTransformer.LoaderSignature)]
[assembly: EntryPointModule(typeof(LpNormNormalization))]
namespace Microsoft.ML.Transforms
{
/// <summary>
/// <see cref="ITransformer"/> resulting from fitting a <see cref="LpNormNormalizingEstimator"/> or <see cref="GlobalContrastNormalizingEstimator"/>.
/// </summary>
public sealed class LpNormNormalizingTransformer : OneToOneTransformerBase
{
internal sealed class Options : TransformInputBase
{
[Argument(ArgumentType.Multiple | ArgumentType.Required, HelpText = "New column definition(s) (optional form: name:src)", Name = "Column", ShortName = "col", SortOrder = 1)]
public Column[] Columns;
[Argument(ArgumentType.AtMostOnce, HelpText = "The norm to use to normalize each sample", ShortName = "norm", SortOrder = 1)]
public LpNormNormalizingEstimatorBase.NormFunction Norm = LpNormNormalizingEstimatorBase.Defaults.Norm;
[Argument(ArgumentType.AtMostOnce, HelpText = "Subtract mean from each value before normalizing", SortOrder = 2)]
public bool SubMean = LpNormNormalizingEstimatorBase.Defaults.LpEnsureZeroMean;
}
internal sealed class GcnOptions : TransformInputBase
{
[Argument(ArgumentType.Multiple, HelpText = "New column definition(s) (optional form: name:src)", Name = "Column", ShortName = "col", SortOrder = 1)]
public GcnColumn[] Columns;
[Argument(ArgumentType.AtMostOnce, HelpText = "Subtract mean from each value before normalizing", SortOrder = 1)]
public bool SubMean = LpNormNormalizingEstimatorBase.Defaults.GcnEnsureZeroMean;
[Argument(ArgumentType.AtMostOnce, HelpText = "Normalize by standard deviation rather than L2 norm", ShortName = "useStd")]
public bool UseStdDev = LpNormNormalizingEstimatorBase.Defaults.EnsureUnitStdDev;
[Argument(ArgumentType.AtMostOnce, HelpText = "Scale features by this value")]
public float Scale = LpNormNormalizingEstimatorBase.Defaults.Scale;
}
internal abstract class ColumnBase : OneToOneColumn
{
[Argument(ArgumentType.AtMostOnce, HelpText = "Subtract mean from each value before normalizing")]
public bool? SubMean;
private protected ColumnBase()
{
}
private protected override bool TryUnparseCore(StringBuilder sb)
{
Contracts.AssertValue(sb);
if (SubMean != null)
return false;
return base.TryUnparseCore(sb);
}
}
internal sealed class Column : ColumnBase
{
[Argument(ArgumentType.AtMostOnce, HelpText = "The norm to use to normalize each sample", ShortName = "norm", SortOrder = 1)]
public LpNormNormalizingEstimatorBase.NormFunction? Norm;
internal static Column Parse(string str)
{
Contracts.AssertNonEmpty(str);
var res = new Column();
if (res.TryParse(str))
return res;
return null;
}
internal bool TryUnparse(StringBuilder sb)
{
Contracts.AssertValue(sb);
if (Norm != null)
return false;
return TryUnparseCore(sb);
}
}
internal sealed class GcnColumn : ColumnBase
{
[Argument(ArgumentType.AtMostOnce, HelpText = "Normalize by standard deviation rather than L2 norm")]
public bool? UseStdDev;
[Argument(ArgumentType.AtMostOnce, HelpText = "Scale features by this value")]
public float? Scale;
internal static GcnColumn Parse(string str)
{
Contracts.AssertNonEmpty(str);
var res = new GcnColumn();
if (res.TryParse(str))
return res;
return null;
}
internal bool TryUnparse(StringBuilder sb)
{
Contracts.AssertValue(sb);
if (UseStdDev != null || Scale != null)
return false;
return TryUnparseCore(sb);
}
}
private sealed class ColumnOptionsLoaded : LpNormNormalizingEstimatorBase.ColumnOptionsBase
{
internal ColumnOptionsLoaded(ModelLoadContext ctx, string name, string inputColumnName, bool normKindSerialized)
: base(ctx, name, inputColumnName, normKindSerialized)
{
}
}
internal const string GcnSummary = "Performs a global contrast normalization on input values: Y = (s * X - M) / D, where s is a scale, M is mean and D is "
+ "either L2 norm or standard deviation.";
internal const string UserNameGn = "Global Contrast Normalization Transform";
internal const string ShortNameGn = "Gcn";
internal const string Summary = "Normalize vectors (rows) individually by rescaling them to unit norm (L2, L1 or LInf). Performs the following operation on a vector X: "
+ "Y = (X - M) / D, where M is mean and D is either L2 norm, L1 norm or LInf norm.";
internal const string UserNameLP = "Lp-Norm Normalizer";
internal const string ShortNameLP = "lpnorm";
private const uint VerVectorNormalizerSupported = 0x00010002;
internal const string LoaderSignature = "GcnTransform";
internal const string LoaderSignatureOld = "GcnFunction";
private static VersionInfo GetVersionInfo()
{
return new VersionInfo(
modelSignature: "GCNORMAF",
// verWrittenCur: 0x00010001, // Initial
// verWrittenCur: 0x00010002, // Added arguments for Lp-norm (vector/row-wise) normalizer
verWrittenCur: 0x00010003, // Dropped sizeof(Float).
verReadableCur: 0x00010003,
verWeCanReadBack: 0x00010001,
loaderSignature: LoaderSignature,
loaderSignatureAlt: LoaderSignatureOld,
loaderAssemblyName: typeof(LpNormNormalizingTransformer).Assembly.FullName);
}
private const string RegistrationName = "LpNormNormalizer";
// REVIEW: should this be an argument instead?
private const float MinScale = 1e-8f;
/// <summary>
/// The objects describing how the transformation is applied on the input data.
/// </summary>
internal IReadOnlyCollection<LpNormNormalizingEstimatorBase.ColumnOptionsBase> Columns => _columns.AsReadOnly();
private readonly LpNormNormalizingEstimatorBase.ColumnOptionsBase[] _columns;
private static (string outputColumnName, string inputColumnName)[] GetColumnPairs(LpNormNormalizingEstimatorBase.ColumnOptionsBase[] columns)
{
Contracts.CheckValue(columns, nameof(columns));
return columns.Select(x => (x.Name, x.InputColumnName)).ToArray();
}
private protected override void CheckInputColumn(DataViewSchema inputSchema, int col, int srcCol)
{
var inType = inputSchema[srcCol].Type;
if (!LpNormNormalizingEstimatorBase.IsColumnTypeValid(inType))
throw Host.ExceptSchemaMismatch(nameof(inputSchema), "input", inputSchema[srcCol].Name, LpNormNormalizingEstimatorBase.ExpectedColumnType, inType.ToString());
}
/// <summary>
/// Create a <see cref="LpNormNormalizingTransformer"/> that takes multiple pairs of columns.
/// </summary>
internal LpNormNormalizingTransformer(IHostEnvironment env, params LpNormNormalizingEstimatorBase.ColumnOptionsBase[] columns) :
base(Contracts.CheckRef(env, nameof(env)).Register(nameof(LpNormNormalizingTransformer)), GetColumnPairs(columns))
{
_columns = columns.ToArray();
}
// Factory method for SignatureDataTransform for GcnArguments class.
internal static IDataTransform Create(IHostEnvironment env, GcnOptions options, IDataView input)
{
Contracts.CheckValue(env, nameof(env));
env.CheckValue(options, nameof(options));
env.CheckValue(input, nameof(input));
env.CheckValue(options.Columns, nameof(options.Columns));
var cols = new GlobalContrastNormalizingEstimator.ColumnOptions[options.Columns.Length];
using (var ch = env.Start("ValidateArgs"))
{
for (int i = 0; i < cols.Length; i++)
{
var item = options.Columns[i];
cols[i] = new GlobalContrastNormalizingEstimator.ColumnOptions(
item.Name,
item.Source ?? item.Name,
item.SubMean ?? options.SubMean,
item.UseStdDev ?? options.UseStdDev,
item.Scale ?? options.Scale);
}
if (!options.SubMean && options.UseStdDev)
ch.Warning("subMean parameter is false while useStd is true. It is advisable to set subMean to true in case useStd is set to true.");
}
return new LpNormNormalizingTransformer(env, cols).MakeDataTransform(input);
}
// Factory method for SignatureDataTransform for Arguments class.
internal static IDataTransform Create(IHostEnvironment env, Options options, IDataView input)
{
Contracts.CheckValue(env, nameof(env));
env.CheckValue(options, nameof(options));
env.CheckValue(input, nameof(input));
env.CheckValue(options.Columns, nameof(options.Columns));
var cols = new LpNormNormalizingEstimator.ColumnOptions[options.Columns.Length];
using (var ch = env.Start("ValidateArgs"))
{
for (int i = 0; i < cols.Length; i++)
{
var item = options.Columns[i];
cols[i] = new LpNormNormalizingEstimator.ColumnOptions(
item.Name,
item.Source ?? item.Name,
item.Norm ?? options.Norm,
item.SubMean ?? options.SubMean);
}
}
return new LpNormNormalizingTransformer(env, cols).MakeDataTransform(input);
}
// Factory method for SignatureLoadModel.
private static LpNormNormalizingTransformer Create(IHostEnvironment env, ModelLoadContext ctx)
{
Contracts.CheckValue(env, nameof(env));
var host = env.Register(nameof(LpNormNormalizingTransformer));
host.CheckValue(ctx, nameof(ctx));
ctx.CheckAtModel(GetVersionInfo());
if (ctx.Header.ModelVerWritten <= 0x00010002)
{
int cbFloat = ctx.Reader.ReadInt32();
env.CheckDecode(cbFloat == sizeof(float));
}
return new LpNormNormalizingTransformer(host, ctx);
}
// Factory method for SignatureLoadDataTransform.
private static IDataTransform Create(IHostEnvironment env, ModelLoadContext ctx, IDataView input)
=> Create(env, ctx).MakeDataTransform(input);
// Factory method for SignatureLoadRowMapper.
private static IRowMapper Create(IHostEnvironment env, ModelLoadContext ctx, DataViewSchema inputSchema)
=> Create(env, ctx).MakeRowMapper(inputSchema);
private LpNormNormalizingTransformer(IHost host, ModelLoadContext ctx)
: base(host, ctx)
{
// *** Binary format ***
// <prefix handled in static Create method>
// <base>
// <columns>
var columnsLength = ColumnPairs.Length;
_columns = new ColumnOptionsLoaded[columnsLength];
for (int i = 0; i < columnsLength; i++)
_columns[i] = new ColumnOptionsLoaded(ctx, ColumnPairs[i].outputColumnName, ColumnPairs[i].inputColumnName, ctx.Header.ModelVerWritten >= VerVectorNormalizerSupported);
}
private protected override void SaveModel(ModelSaveContext ctx)
{
Host.CheckValue(ctx, nameof(ctx));
ctx.CheckAtModel();
ctx.SetVersionInfo(GetVersionInfo());
// *** Binary format ***
// <base>
// <columns>
SaveColumns(ctx);
Host.Assert(_columns.Length == ColumnPairs.Length);
foreach (var col in _columns)
col.Save(ctx);
}
private protected override IRowMapper MakeRowMapper(DataViewSchema schema) => new Mapper(this, schema);
private sealed class Mapper : OneToOneMapperBase, ISaveAsOnnx
{
private readonly DataViewType[] _srcTypes;
private readonly int[] _srcCols;
private readonly DataViewType[] _types;
private readonly LpNormNormalizingEstimatorBase.NormFunction[] _norms;
private readonly bool[] _ensureZeroMeans;
private readonly LpNormNormalizingTransformer _parent;
public Mapper(LpNormNormalizingTransformer parent, DataViewSchema inputSchema)
: base(parent.Host.Register(nameof(Mapper)), parent, inputSchema)
{
_parent = parent;
_types = new DataViewType[_parent.ColumnPairs.Length];
_srcTypes = new DataViewType[_parent.ColumnPairs.Length];
_srcCols = new int[_parent.ColumnPairs.Length];
_norms = new LpNormNormalizingEstimatorBase.NormFunction[_parent.ColumnPairs.Length];
_ensureZeroMeans = new bool[_parent.ColumnPairs.Length];
for (int i = 0; i < _parent.ColumnPairs.Length; i++)
{
inputSchema.TryGetColumnIndex(_parent.ColumnPairs[i].inputColumnName, out _srcCols[i]);
var srcCol = inputSchema[_srcCols[i]];
_srcTypes[i] = srcCol.Type;
_types[i] = srcCol.Type;
_norms[i] = _parent._columns[i].Norm;
_ensureZeroMeans[i] = _parent._columns[i].EnsureZeroMean;
}
}
protected override DataViewSchema.DetachedColumn[] GetOutputColumnsCore()
{
var result = new DataViewSchema.DetachedColumn[_parent.ColumnPairs.Length];
for (int i = 0; i < _parent.ColumnPairs.Length; i++)
{
var builder = new DataViewSchema.Annotations.Builder();
builder.Add(InputSchema[ColMapNewToOld[i]].Annotations, name => name == AnnotationUtils.Kinds.SlotNames);
ValueGetter<bool> getter = (ref bool dst) => dst = true;
builder.Add(AnnotationUtils.Kinds.IsNormalized, BooleanDataViewType.Instance, getter);
result[i] = new DataViewSchema.DetachedColumn(_parent.ColumnPairs[i].outputColumnName, _types[i], builder.ToAnnotations());
}
return result;
}
protected override Delegate MakeGetter(DataViewRow input, int iinfo, Func<int, bool> activeOutput, out Action disposer)
{
Contracts.AssertValue(input);
Contracts.Assert(0 <= iinfo && iinfo < _parent.ColumnPairs.Length);
disposer = null;
var ex = _parent._columns[iinfo];
Host.Assert(0 < ex.Scale && ex.Scale < float.PositiveInfinity);
Host.Assert(_srcTypes[iinfo] is VectorDataViewType);
var getSrc = input.GetGetter<VBuffer<float>>(input.Schema[_srcCols[iinfo]]);
var src = default(VBuffer<float>);
ValueGetter<VBuffer<float>> del;
float scale = ex.Scale;
if (ex.EnsureZeroMean)
{
switch (ex.Norm)
{
case LpNormNormalizingEstimatorBase.NormFunction.StandardDeviation:
del =
(ref VBuffer<float> dst) =>
{
getSrc(ref src);
var srcValues = src.GetValues();
var mean = Mean(srcValues, src.Length);
var divisor = StdDev(srcValues, src.Length, mean);
FillValues(Host, in src, ref dst, divisor, scale, mean);
};
return del;
case LpNormNormalizingEstimatorBase.NormFunction.L2:
del =
(ref VBuffer<float> dst) =>
{
getSrc(ref src);
var srcValues = src.GetValues();
var mean = Mean(srcValues, src.Length);
var divisor = L2Norm(srcValues, mean);
FillValues(Host, in src, ref dst, divisor, scale, mean);
};
return del;
case LpNormNormalizingEstimatorBase.NormFunction.L1:
del =
(ref VBuffer<float> dst) =>
{
getSrc(ref src);
var srcValues = src.GetValues();
var mean = Mean(srcValues, src.Length);
var divisor = L1Norm(srcValues, mean);
FillValues(Host, in src, ref dst, divisor, scale, mean);
};
return del;
case LpNormNormalizingEstimatorBase.NormFunction.Infinity:
del =
(ref VBuffer<float> dst) =>
{
getSrc(ref src);
var srcValues = src.GetValues();
var mean = Mean(srcValues, src.Length);
var divisor = LInfNorm(srcValues, mean);
FillValues(Host, in src, ref dst, divisor, scale, mean);
};
return del;
default:
Host.Assert(false, "Unsupported normalizer type");
goto case LpNormNormalizingEstimatorBase.NormFunction.L2;
}
}
switch (ex.Norm)
{
case LpNormNormalizingEstimatorBase.NormFunction.StandardDeviation:
del =
(ref VBuffer<float> dst) =>
{
getSrc(ref src);
var divisor = StdDev(src.GetValues(), src.Length);
FillValues(Host, in src, ref dst, divisor, scale);
};
return del;
case LpNormNormalizingEstimatorBase.NormFunction.L2:
del =
(ref VBuffer<float> dst) =>
{
getSrc(ref src);
var divisor = L2Norm(src.GetValues());
FillValues(Host, in src, ref dst, divisor, scale);
};
return del;
case LpNormNormalizingEstimatorBase.NormFunction.L1:
del =
(ref VBuffer<float> dst) =>
{
getSrc(ref src);
var divisor = L1Norm(src.GetValues());
FillValues(Host, in src, ref dst, divisor, scale);
};
return del;
case LpNormNormalizingEstimatorBase.NormFunction.Infinity:
del =
(ref VBuffer<float> dst) =>
{
getSrc(ref src);
var divisor = LInfNorm(src.GetValues());
FillValues(Host, in src, ref dst, divisor, scale);
};
return del;
default:
Host.Assert(false, "Unsupported normalizer type");
goto case LpNormNormalizingEstimatorBase.NormFunction.L2;
}
}
private static void FillValues(IExceptionContext ectx, in VBuffer<float> src, ref VBuffer<float> dst, float divisor, float scale, float offset = 0)
{
var srcValues = src.GetValues();
int count = srcValues.Length;
int length = src.Length;
ectx.Assert(divisor >= 0);
if (count == 0)
{
VBufferUtils.Resize(ref dst, length, 0);
return;
}
ectx.Assert(count > 0);
ectx.Assert(length > 0);
float normScale = scale;
if (divisor > 0)
normScale /= divisor;
// Don't normalize small values.
if (normScale < MinScale)
normScale = 1;
VBufferEditor<float> editor;
if (offset == 0)
{
editor = VBufferEditor.Create(ref dst, length, count);
var dstValues = editor.Values;
if (!src.IsDense)
{
src.GetIndices().CopyTo(editor.Indices);
}
CpuMathUtils.Scale(normScale, src.GetValues(), dstValues, count);
dst = editor.Commit();
return;
}
// Subtracting the mean requires a dense representation.
src.CopyToDense(ref dst);
editor = VBufferEditor.CreateFromBuffer(ref dst);
if (normScale != 1)
CpuMathUtils.ScaleAdd(normScale, -offset, editor.Values);
else
CpuMathUtils.Add(-offset, editor.Values);
}
/// <summary>
/// Compute Standard Deviation. In case of both subMean and useStd are true, we technically need to compute variance
/// based on centered values (i.e. after subtracting the mean). But since the centered
/// values mean is approximately zero, we can use variance of non-centered values.
/// </summary>
private static float StdDev(ReadOnlySpan<float> values, int length)
{
Contracts.Assert(0 <= values.Length && values.Length <= length);
if (values.Length == 0)
return 0;
// We need a mean to compute variance.
var tmpMean = CpuMathUtils.Sum(values) / length;
float sumSq = 0;
if (values.Length != length && tmpMean != 0)
{
// Sparse representation.
float meanSq = tmpMean * tmpMean;
sumSq = (length - values.Length) * meanSq;
}
sumSq += CpuMathUtils.SumSq(tmpMean, values);
return MathUtils.Sqrt(sumSq / length);
}
/// <summary>
/// Compute Standard Deviation.
/// We have two overloads of StdDev instead of one with <see cref="Nullable{Float}"/> mean for perf reasons.
/// </summary>
private static float StdDev(ReadOnlySpan<float> values, int length, float mean)
{
Contracts.Assert(0 <= values.Length && values.Length <= length);
if (values.Length == 0)
return 0;
float sumSq = 0;
if (values.Length != length && mean != 0)
{
// Sparse representation.
float meanSq = mean * mean;
sumSq = (length - values.Length) * meanSq;
}
sumSq += CpuMathUtils.SumSq(mean, values);
return MathUtils.Sqrt(sumSq / length);
}
/// <summary>
/// Compute L2-norm. L2-norm computation doesn't subtract the mean from the source values.
/// However, we substract the mean here in case subMean is true (if subMean is false, mean is zero).
/// </summary>
private static float L2Norm(ReadOnlySpan<float> values, float mean = 0)
{
if (values.Length == 0)
return 0;
return MathUtils.Sqrt(CpuMathUtils.SumSq(mean, values));
}
/// <summary>
/// Compute L1-norm. L1-norm computation doesn't subtract the mean from the source values.
/// However, we substract the mean here in case subMean is true (if subMean is false, mean is zero).
/// </summary>
private static float L1Norm(ReadOnlySpan<float> values, float mean = 0)
{
if (values.Length == 0)
return 0;
return CpuMathUtils.SumAbs(mean, values);
}
/// <summary>
/// Compute LInf-norm. LInf-norm computation doesn't subtract the mean from the source values.
/// However, we substract the mean here in case subMean is true (if subMean is false, mean is zero).
/// </summary>
private static float LInfNorm(ReadOnlySpan<float> values, float mean = 0)
{
if (values.Length == 0)
return 0;
return CpuMathUtils.MaxAbsDiff(mean, values);
}
private static float Mean(ReadOnlySpan<float> src, int length)
{
if (length == 0 || src.Length == 0)
return 0;
return CpuMathUtils.Sum(src) / length;
}
public bool CanSaveOnnx(OnnxContext ctx) => true;
public void SaveAsOnnx(OnnxContext ctx)
{
Host.CheckValue(ctx, nameof(ctx));
for (int iinfo = 0; iinfo < _srcCols.Length; ++iinfo)
{
string inputColumnName = InputSchema[_srcCols[iinfo]].Name;
if (!ctx.ContainsColumn(inputColumnName))
{
ctx.RemoveColumn(inputColumnName, false);
continue;
}
string outputColumnName = _parent.ColumnPairs[iinfo].outputColumnName;
if (!SaveAsOnnxCore(ctx, iinfo, ctx.GetVariableName(inputColumnName), ctx.AddIntermediateVariable(_srcTypes[iinfo], outputColumnName)))
{
ctx.RemoveColumn(inputColumnName, true);
}
}
}
private bool SaveAsOnnxCore(OnnxContext ctx, int iinfo, string srcVariableName, string dstVariableName)
{
const int minimumOpSetVersion = 9;
ctx.CheckOpSetVersion(minimumOpSetVersion, LoaderSignature);
string opType;
if ((_norms[iinfo] != LpNormNormalizingEstimatorBase.NormFunction.StandardDeviation) && (_ensureZeroMeans[iinfo] == false))
{
string strNorm;
if (_norms[iinfo] == LpNormNormalizingEstimatorBase.NormFunction.L1)
strNorm = "L1";
else if (_norms[iinfo] == LpNormNormalizingEstimatorBase.NormFunction.L2)
strNorm = "L2";
else
strNorm = "MAX";
opType = "Normalizer";
var node = ctx.CreateNode(opType, srcVariableName, dstVariableName, ctx.GetNodeName(opType));
node.AddAttribute("norm", strNorm);
return true;
}
opType = "ReduceMean";
string meanOfInput = ctx.AddIntermediateVariable(_types[iinfo], "MeanOfInput", true);
var meanNode = ctx.CreateNode(opType, srcVariableName, meanOfInput, ctx.GetNodeName(opType), "");
meanNode.AddAttribute("axes", new long[] { 1 });
opType = "Sub";
string inputMinusMean = ctx.AddIntermediateVariable(_types[iinfo], "InputMinusMean");
var subtractNode = ctx.CreateNode(opType, new[] { srcVariableName, meanOfInput }, new[] { inputMinusMean }, ctx.GetNodeName(opType), "");
if (_norms[iinfo] == LpNormNormalizingEstimatorBase.NormFunction.L1)
{
opType = "Abs";
string absOfInput = ctx.AddIntermediateVariable(_types[iinfo], "AbsOfInput");
var absNode = ctx.CreateNode(opType, inputMinusMean, absOfInput, ctx.GetNodeName(opType), "");
opType = "ReduceSum";
string sumOfAbsOfInput = ctx.AddIntermediateVariable(_types[iinfo], "SumOfAbsOfInput", true);
var sumOfAbsNode = ctx.CreateNode(opType, absOfInput, sumOfAbsOfInput, ctx.GetNodeName(opType), "");
sumOfAbsNode.AddAttribute("axes", new long[] { 1 });
opType = "Div";
var l1Node = ctx.CreateNode(opType, new[] { inputMinusMean, sumOfAbsOfInput }, new[] { dstVariableName }, ctx.GetNodeName(opType), "");
}
else if (_norms[iinfo] == LpNormNormalizingEstimatorBase.NormFunction.L2)
{
opType = "Pow";
string two = ctx.AddInitializer(2.0f);
string squareOfInput = ctx.AddIntermediateVariable(_types[iinfo], "SquareOfInput", true);
var squareNode = ctx.CreateNode(opType, new[] { inputMinusMean, two }, new[] { squareOfInput }, ctx.GetNodeName(opType), "");
opType = "ReduceSum";
string sumOfSquares = ctx.AddIntermediateVariable(_types[iinfo], "SumOfSquares", true);
var sumOfSquaresNode = ctx.CreateNode(opType, squareOfInput, sumOfSquares, ctx.GetNodeName(opType), "");
sumOfSquaresNode.AddAttribute("axes", new long[] { 1 });
opType = "Sqrt";
string squareRoot = ctx.AddIntermediateVariable(_types[iinfo], "SquareRoot", true);
var squareRootNode = ctx.CreateNode(opType, sumOfSquares, squareRoot, ctx.GetNodeName(opType), "");
opType = "Div";
var l2Node = ctx.CreateNode(opType, new[] { inputMinusMean, squareRoot }, new[] { dstVariableName }, ctx.GetNodeName(opType), "");
}
else if (_norms[iinfo] == LpNormNormalizingEstimatorBase.NormFunction.Infinity)
{
opType = "ReduceMax";
string maxOfInput = ctx.AddIntermediateVariable(_types[iinfo], "MaxOfInput", true);
var maxNode = ctx.CreateNode(opType, inputMinusMean, maxOfInput, ctx.GetNodeName(opType), "");
maxNode.AddAttribute("axes", new long[] { 1 });
opType = "Div";
var lMaxNode = ctx.CreateNode(opType, new[] { inputMinusMean, maxOfInput }, new[] { dstVariableName }, ctx.GetNodeName(opType), "");
}
else if (_norms[iinfo] == LpNormNormalizingEstimatorBase.NormFunction.StandardDeviation)
{
// first calculate the standard deviation
opType = "Pow";
string two = ctx.AddInitializer(2.0f);
string squareOfInputMinusMean = ctx.AddIntermediateVariable(_types[iinfo], "SquareOfInputMinusMean", true);
var squareOfInputMinusMeanNode = ctx.CreateNode(opType, new[] { inputMinusMean, two }, new[] { squareOfInputMinusMean }, ctx.GetNodeName(opType), "");
opType = "ReduceMean";
string average = ctx.AddIntermediateVariable(_types[iinfo], "SumOfSquares", true);
var sumOfSquaresNode = ctx.CreateNode(opType, squareOfInputMinusMean, average, ctx.GetNodeName(opType), "");
sumOfSquaresNode.AddAttribute("axes", new long[] { 1 });
opType = "Sqrt";
string stdDev = ctx.AddIntermediateVariable(_types[iinfo], "SquareRoot", true);
var stdDevNode = ctx.CreateNode(opType, average, stdDev, ctx.GetNodeName(opType), "");
opType = "Div";
string input = _ensureZeroMeans[iinfo] ? inputMinusMean : srcVariableName;
var lStdDevNode = ctx.CreateNode(opType, new[] {input, stdDev }, new[] { dstVariableName }, ctx.GetNodeName(opType), "");
}
else
{
Contracts.Assert(false);
return false;
}
return true;
}
}
}
internal static class LpNormNormalization
{
[TlcModule.EntryPoint(Name = "Transforms.LpNormalizer",
Desc = LpNormNormalizingTransformer.Summary,
UserName = LpNormNormalizingTransformer.UserNameLP,
ShortName = LpNormNormalizingTransformer.ShortNameLP)]
public static CommonOutputs.TransformOutput Normalize(IHostEnvironment env, LpNormNormalizingTransformer.Options input)
{
var h = EntryPointUtils.CheckArgsAndCreateHost(env, "LpNormalize", input);
var xf = LpNormNormalizingTransformer.Create(h, input, input.Data);
return new CommonOutputs.TransformOutput()
{
Model = new TransformModelImpl(h, xf, input.Data),
OutputData = xf
};
}
[TlcModule.EntryPoint(Name = "Transforms.GlobalContrastNormalizer",
Desc = LpNormNormalizingTransformer.GcnSummary,
UserName = LpNormNormalizingTransformer.UserNameGn,
ShortName = LpNormNormalizingTransformer.ShortNameGn)]
public static CommonOutputs.TransformOutput GcNormalize(IHostEnvironment env, LpNormNormalizingTransformer.GcnOptions input)
{
var h = EntryPointUtils.CheckArgsAndCreateHost(env, "GcNormalize", input);
var xf = LpNormNormalizingTransformer.Create(h, input, input.Data);
return new CommonOutputs.TransformOutput()
{
Model = new TransformModelImpl(h, xf, input.Data),
OutputData = xf
};
}
}
/// <summary>
/// Base estimator class for <see cref="LpNormNormalizingEstimator"/> and <see cref="GlobalContrastNormalizingEstimator"/> normalizers.
/// </summary>
public abstract class LpNormNormalizingEstimatorBase : TrivialEstimator<LpNormNormalizingTransformer>
{
/// <summary>
/// The kind of unit norm vectors are rescaled to. This enumeration is serialized.
/// </summary>
public enum NormFunction : byte
{
/// <summary>
/// L2-norm.
/// </summary>
L2 = 0,
/// <summary>
/// Standard deviation of a vector by viewing all its coordinates as a random variable.
/// </summary>
StandardDeviation = 1,
/// <summary>
/// L1-norm.
/// </summary>
L1 = 2,
/// <summary>
/// Infinity-norm.
/// </summary>
Infinity = 3
}
/// <summary>
/// Describes base class for one column pair.
/// </summary>
internal abstract class ColumnOptionsBase
{
/// <summary>
/// Name of the column resulting from the transformation of <see cref="InputColumnName"/>.
/// </summary>
public readonly string Name;
/// <summary>
/// Name of column to transform.
/// </summary>
public readonly string InputColumnName;
/// <summary>
/// The norm to use to normalize each sample.
/// </summary>
public readonly NormFunction Norm;
/// <summary>
/// Subtract mean from each value before normalizing.
/// </summary>
public readonly bool EnsureZeroMean;
/// <summary>
/// Scale features by this value.
/// </summary>
public readonly float Scale;
internal ColumnOptionsBase(string name, string inputColumnName, NormFunction normalizerKind, bool substractMean, float scale)
{
Contracts.CheckNonWhiteSpace(name, nameof(name));
Contracts.CheckNonWhiteSpace(inputColumnName, nameof(inputColumnName));
Name = name;
InputColumnName = inputColumnName;
EnsureZeroMean = substractMean;
Contracts.CheckUserArg(0 < scale && scale < float.PositiveInfinity, nameof(scale), "scale must be a positive finite value");
Scale = scale;
Norm = normalizerKind;
}
internal ColumnOptionsBase(ModelLoadContext ctx, string name, string inputColumnName, bool normKindSerialized)
{
Contracts.AssertValue(ctx);
Contracts.CheckNonWhiteSpace(inputColumnName, nameof(inputColumnName));
Contracts.CheckNonWhiteSpace(name, nameof(name));
Name = name;
InputColumnName = inputColumnName;
// *** Binary format ***
// byte: SubtractMean
// byte: NormKind
// Float: Scale
EnsureZeroMean = ctx.Reader.ReadBoolByte();
byte normKindVal = ctx.Reader.ReadByte();
Contracts.CheckDecode(Enum.IsDefined(typeof(NormFunction), normKindVal));
Norm = (NormFunction)normKindVal;
// Note: In early versions, a bool option (useStd) to whether to normalize by StdDev rather than
// L2 norm was used. normKind was added in version=verVectorNormalizerSupported.
// normKind was defined in a way such that the serialized boolean (0: use StdDev, 1: use L2) is
// still valid.
Contracts.CheckDecode(normKindSerialized ||
(Norm == NormFunction.L2 || Norm == NormFunction.StandardDeviation));
Scale = ctx.Reader.ReadFloat();
Contracts.CheckDecode(0 < Scale && Scale < float.PositiveInfinity);
}
internal void Save(ModelSaveContext ctx)
{
Contracts.AssertValue(ctx);
// *** Binary format ***
// byte: SubtractMean
// byte: NormKind
// Float: Scale
ctx.Writer.WriteBoolByte(EnsureZeroMean);
ctx.Writer.Write((byte)Norm);
Contracts.Assert(0 < Scale && Scale < float.PositiveInfinity);
ctx.Writer.Write(Scale);
}
}
[BestFriend]
internal static class Defaults
{
public const NormFunction Norm = NormFunction.L2;
public const bool LpEnsureZeroMean = false;
public const bool GcnEnsureZeroMean = true;
public const bool EnsureUnitStdDev = false;
public const float Scale = 1;
}
/// <summary>
/// Create a <see cref="LpNormNormalizingEstimatorBase"/> that takes multiple pairs of columns.
/// </summary>
internal LpNormNormalizingEstimatorBase(IHostEnvironment env, params ColumnOptionsBase[] columns)
: base(Contracts.CheckRef(env, nameof(env)).Register(nameof(LpNormNormalizingEstimator)), new LpNormNormalizingTransformer(env, columns))
{
}
internal static bool IsColumnTypeValid(DataViewType type)
{
if (!(type is VectorDataViewType vectorType && vectorType.IsKnownSize))
return false;
return vectorType.ItemType == NumberDataViewType.Single;
}
internal static bool IsSchemaColumnValid(SchemaShape.Column col)
{
if (col.Kind != SchemaShape.Column.VectorKind.Vector)
return false;
return col.ItemType == NumberDataViewType.Single;
}
internal const string ExpectedColumnType = "Expected known-size vector of Single";
/// <summary>
/// Returns the <see cref="SchemaShape"/> of the schema which will be produced by the transformer.
/// Used for schema propagation and verification in a pipeline.
/// </summary>
public override SchemaShape GetOutputSchema(SchemaShape inputSchema)
{
Host.CheckValue(inputSchema, nameof(inputSchema));
var result = inputSchema.ToDictionary(x => x.Name);
foreach (var colPair in Transformer.Columns)
{
if (!inputSchema.TryFindColumn(colPair.InputColumnName, out var col))
throw Host.ExceptSchemaMismatch(nameof(inputSchema), "input", colPair.InputColumnName);
if (!IsSchemaColumnValid(col))
throw Host.ExceptSchemaMismatch(nameof(inputSchema), "input", colPair.InputColumnName, ExpectedColumnType, col.GetTypeString());
var metadata = new List<SchemaShape.Column>();
if (col.Annotations.TryFindColumn(AnnotationUtils.Kinds.SlotNames, out var slotMeta))
metadata.Add(slotMeta);
metadata.Add(new SchemaShape.Column(AnnotationUtils.Kinds.IsNormalized, SchemaShape.Column.VectorKind.Scalar, BooleanDataViewType.Instance, false));
result[colPair.Name] = new SchemaShape.Column(colPair.Name, col.Kind, col.ItemType, false, new SchemaShape(metadata.ToArray()));
}
return new SchemaShape(result.Values);
}
}
/// <summary>
/// Normalizes (scales) vectors in the input column to the unit norm. The type of norm that is used can be specified by the user.
/// </summary>
/// <remarks>
/// <format type="text/markdown"><![CDATA[
///
/// ### Estimator Characteristics
/// | | |
/// | -- | -- |
/// | Does this estimator need to look at the data to train its parameters? | No |
/// | Input column data type | Vector of <xref:System.Single> |
/// | Output column data type | Vector of <xref:System.Single> |
/// | Exportable to ONNX | Yes |
///
/// The resulting <xref:Microsoft.ML.Transforms.LpNormNormalizingTransformer> normalizes vectors in the input column individually
/// by rescaling them to the unit norm.
///
/// Let $x$ be the input vector, $n$ the size of the vector, $L(x)$ the norm function selected by the user.
/// Let $\mu(x) = \sum_i x_i / n$ be the mean of the values of vector $x$. The <xref:Microsoft.ML.Transforms.LpNormNormalizingTransformer>
/// performs the following operation on each input vector $x$:
/// $y = \frac{x - \mu(x)}{L(x)}$
/// if the user specifies that the mean should be zero, or otherwise:
/// $y = \frac{x}{L(x)}$
///
/// There are four types of norm that can be selected by the user to be applied on input vector $x$. They are defined as follows:
/// - <xref:Microsoft.ML.Transforms.LpNormNormalizingEstimatorBase.NormFunction.L1>: $L_1(x) = \sum_i |x_i|$
/// - <xref:Microsoft.ML.Transforms.LpNormNormalizingEstimatorBase.NormFunction.L2>: $L_2(x) = \sqrt{\sum_i x_i^2}$
/// - <xref:Microsoft.ML.Transforms.LpNormNormalizingEstimatorBase.NormFunction.Infinity>: $L_{\infty}(x) = \max_i\{|x_i|\}$
/// - <xref:Microsoft.ML.Transforms.LpNormNormalizingEstimatorBase.NormFunction.StandardDeviation>: $L_\sigma(x)$ is defined as the
/// standard deviation of the elements of the input vector $x$
///
/// Check the See Also section for links to usage examples.
/// ]]>
/// </format>
/// </remarks>
/// <seealso cref="NormalizationCatalog.NormalizeLpNorm(TransformsCatalog, string, string, LpNormNormalizingEstimatorBase.NormFunction, bool)"/>
public sealed class LpNormNormalizingEstimator : LpNormNormalizingEstimatorBase
{
/// <summary>
/// Describes how the transformer handles one column pair.
/// </summary>
internal sealed class ColumnOptions : ColumnOptionsBase
{
/// <summary>
/// Describes how the transformer handles one column pair.
/// </summary>
/// <param name="name">Name of the column resulting from the transformation of <paramref name="inputColumnName"/>.</param>
/// <param name="inputColumnName">Name of column to transform. If set to <see langword="null"/>, the value of the <paramref name="name"/> will be used as source.</param>
/// <param name="norm">Type of norm to use to normalize each sample. The indicated norm of the resulted vector will be normalized to one.</param>
/// <param name="ensureZeroMean">If <see langword="true"/>, subtract mean from each value before normalizing and use the raw input otherwise.</param>
public ColumnOptions(string name, string inputColumnName = null,
NormFunction norm = Defaults.Norm,
bool ensureZeroMean = Defaults.LpEnsureZeroMean)
: base(name, inputColumnName ?? name, norm, ensureZeroMean, 1)
{
}
}
/// <include file='doc.xml' path='doc/members/member[@name="LpNormalize"]/*'/>
/// <param name="env">The environment.</param>
/// <param name="outputColumnName">Name of the column resulting from the transformation of <paramref name="inputColumnName"/>.</param>
/// <param name="inputColumnName">Name of the column to transform. If set to <see langword="null"/>, the value of the <paramref name="outputColumnName"/> will be used as source.</param>
/// <param name="norm">Type of norm to use to normalize each sample. The indicated norm of the resulted vector will be normalized to one.</param>
/// <param name="ensureZeroMean">If <see langword="true"/>, subtract mean from each value before normalizing and use the raw input otherwise.</param>
internal LpNormNormalizingEstimator(IHostEnvironment env, string outputColumnName, string inputColumnName = null,
NormFunction norm = Defaults.Norm, bool ensureZeroMean = Defaults.LpEnsureZeroMean)
: this(env, new[] { (outputColumnName, inputColumnName ?? outputColumnName) }, norm, ensureZeroMean)
{
}
/// <include file='doc.xml' path='doc/members/member[@name="LpNormalize"]/*'/>
/// <param name="env">The environment.</param>