Thanks to visit codestin.com
Credit goes to github.com

Skip to content

Commit f2637e7

Browse files
deepakkumar1984deepakkumar1984
committed
Updating Gluon code to use numpy array instead of MxNet NDAray which is being deprecated
1 parent 0da2dbb commit f2637e7

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

72 files changed

+675
-357
lines changed

‎examples/BasicExamples/Program.cs‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -12,7 +12,7 @@ class Program
1212
static void Main(string[] args)
1313
{
1414
//Console.WriteLine("Runnin XOR Example......");
15-
//XORGate.Run();
15+
XORGate.Run();
1616
//CrashCourse_NN.Run();
1717
//LogisticRegressionExplained.Run();
1818
//var methods = mx.GetAllRegisteredCApiOperators();

‎examples/BasicExamples/XORGate.cs‎

Lines changed: 7 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -8,6 +8,7 @@
88
using MxNet.Optimizers;
99
using System;
1010
using System.Linq;
11+
using MxNet.Numpy;
1112

1213
namespace BasicExamples
1314
{
@@ -16,8 +17,8 @@ public class XORGate
1617
public static void Run()
1718
{
1819
// Create
19-
var trainX = new NDArray(new float[] { 0, 0, 0, 1, 1, 0, 1, 1 }).Reshape(4, 2);
20-
var trainY = new NDArray(new float[] { 0, 1, 1, 0 });
20+
var trainX = new ndarray(new float[] { 0, 0, 0, 1, 1, 0, 1, 1 }).reshape(new Shape(4, 2));
21+
var trainY = new ndarray(new float[] { 0, 1, 1, 0 });
2122

2223
var batch_size = 2;
2324
var train_data = new NDArrayIter(trainX, trainY, batch_size);
@@ -43,17 +44,17 @@ public static void Run()
4344
while (!train_data.End())
4445
{
4546
var batch = train_data.Next();
46-
var data = Utils.SplitAndLoad(batch.Data[0], ctxList);
47-
var label = Utils.SplitAndLoad(batch.Label[0], ctxList);
47+
var data = MxNet.Gluon.Utils.SplitAndLoad(batch.Data[0], ctxList);
48+
var label = MxNet.Gluon.Utils.SplitAndLoad(batch.Label[0], ctxList);
4849
NDArrayList outputs = null;
4950
using (var ag = Autograd.Record())
5051
{
5152
outputs = Enumerable.Zip(data, label, (x, y) =>
5253
{
5354
var z = net.Call(x);
54-
NDArray loss = binary_crossentropy.Call(z, y);
55+
ndarray loss = binary_crossentropy.Call(z, y);
5556
loss.Backward();
56-
lossVal += loss.Mean();
57+
lossVal += loss.mean().AsScalar<float>();
5758
return z;
5859
}).ToList();
5960
}

‎src/MxNet/AMP/Amp.cs‎

Lines changed: 0 additions & 10 deletions
Original file line numberDiff line numberDiff line change
@@ -10,15 +10,5 @@ public static NDArrayOrSymbol CastSymbolNDArray(NDArrayOrSymbol s, DType dtype,
1010
{
1111
throw new NotImplementedException();
1212
}
13-
14-
public static (string, Type[]) GetNdFuncToWrap(string name, Type module, Dictionary<string, object> submodule_dict)
15-
{
16-
throw new NotImplementedException();
17-
}
18-
19-
public static (string, Type[]) GetNpFuncToWrap(string name, Type module, Dictionary<string, object> submodule_dict)
20-
{
21-
throw new NotImplementedException();
22-
}
2313
}
2414
}

‎src/MxNet/F.cs‎

Lines changed: 15 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -112,7 +112,7 @@ public static NDArrayOrSymbol cumsum(NDArrayOrSymbol a, int? axis, DType dtype,
112112
}
113113
return sym_np_ops.cumsum(a, axis, dtype, @out);
114114
}
115-
public static NDArrayOrSymbol reshape(NDArrayOrSymbol a, Shape newshape, bool reverse, String order)
115+
public static NDArrayOrSymbol reshape(NDArrayOrSymbol a, Shape newshape, bool reverse = false, String order = "C")
116116
{
117117
if (a.IsNDArray)
118118
{
@@ -176,7 +176,7 @@ public static NDArrayOrSymbol diagonal(NDArrayOrSymbol a, int offset, int axis1,
176176
}
177177
return sym_np_ops.diagonal(a, offset, axis1, axis2);
178178
}
179-
public static NDArrayOrSymbol sum(NDArrayOrSymbol a, int? axis, DType dtype, NDArrayOrSymbol @out, bool keepdims, float? initial)
179+
public static NDArrayOrSymbol sum(NDArrayOrSymbol a, int? axis = null, DType dtype = null, NDArrayOrSymbol @out = null, bool keepdims = false, float? initial = null)
180180
{
181181
if (a.IsNDArray)
182182
{
@@ -328,7 +328,7 @@ public static NDArrayOrSymbol softmax(NDArrayOrSymbol data, int axis, NDArrayOrS
328328
}
329329
return Sym.Numpy.npx.softmax(data, axis, length, temperature, use_length, dtype);
330330
}
331-
public static NDArrayOrSymbol log_softmax(NDArrayOrSymbol data, int axis, NDArrayOrSymbol length, Double? temperature, bool use_length, DType dtype)
331+
public static NDArrayOrSymbol log_softmax(NDArrayOrSymbol data, int axis, NDArrayOrSymbol length = null, Double? temperature = null, bool use_length = false, DType dtype = null)
332332
{
333333
if (data.IsNDArray)
334334
{
@@ -360,7 +360,7 @@ public static NDArrayOrSymbol pick(NDArrayOrSymbol data, NDArrayOrSymbol index,
360360
}
361361
return Sym.Numpy.npx.pick(data, index, axis, mode, keepdims);
362362
}
363-
public static NDArrayOrSymbol reshape_like(NDArrayOrSymbol lhs, NDArrayOrSymbol rhs, int? lhs_begin, int? lhs_end, int? rhs_begin, int? rhs_end)
363+
public static NDArrayOrSymbol reshape_like(NDArrayOrSymbol lhs, NDArrayOrSymbol rhs, int? lhs_begin = null, int? lhs_end = null, int? rhs_begin = null, int? rhs_end = null)
364364
{
365365
if (lhs.IsNDArray)
366366
{
@@ -563,7 +563,7 @@ public static NDArrayOrSymbol average(NDArrayOrSymbol a, int? axis, NDArrayOrSym
563563
}
564564
return sym_np_ops.average(a, axis, weights, returned, @out);
565565
}
566-
public static NDArrayOrSymbol mean(NDArrayOrSymbol a, int? axis, DType dtype, NDArrayOrSymbol @out, bool keepdims)
566+
public static NDArrayOrSymbol mean(NDArrayOrSymbol a, int? axis = null, DType dtype = null, NDArrayOrSymbol @out = null, bool keepdims = false)
567567
{
568568
if (a.IsNDArray)
569569
{
@@ -1575,5 +1575,15 @@ public static NDArrayOrSymbolList split(NDArrayOrSymbol ary, int[] indices_or_se
15751575
}
15761576
return sym_np_ops.split(ary, indices_or_sections, axis);
15771577
}
1578+
1579+
public static NDArrayOrSymbol norm(NDArrayOrSymbol x, string ord = null, Shape axis = null, bool keepdims = false)
1580+
{
1581+
if (x.IsNDArray)
1582+
{
1583+
return nd_np_ops.linalg.norm(x, ord, axis, keepdims);
1584+
}
1585+
1586+
return sym_np_ops.linalg.norm(x, ord, axis, keepdims);
1587+
}
15781588
}
15791589
}

‎src/MxNet/Gluon/Block/Block.cs‎

Lines changed: 2 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -50,6 +50,7 @@ public Block()
5050
_reg_params = new Dictionary<string, Parameter>();
5151
_forward_hooks = new Dictionary<int, Hook>();
5252
_forward_pre_hooks = new Dictionary<int, Hook>();
53+
Params = new ParameterDict();
5354
}
5455

5556
public virtual ParameterDict Params { get; set; }
@@ -591,7 +592,7 @@ internal static string CommonPrefix(string[] names)
591592
return prefix;
592593
}
593594

594-
internal static (DType[], DType[]) InferParamTypes(SymbolList in_params, Symbol out_params, string[] arg_params,
595+
internal static (DType[], DType[]) InferParamTypes(SymbolList in_params, _Symbol out_params, string[] arg_params,
595596
string[] aux_params, DType default_dtype = null)
596597
{
597598
DType[] arg_types = null;

‎src/MxNet/Gluon/Block/HybridBlock.cs‎

Lines changed: 8 additions & 8 deletions
Original file line numberDiff line numberDiff line change
@@ -371,7 +371,7 @@ private void BuildCache(NDArrayOrSymbolList args, bool update_graph = true)
371371
for (var i = 0; i < input_names.Length; i++)
372372
{
373373
Parameter param = null;
374-
NDArray param_data;
374+
ndarray param_data;
375375
var name = input_names[i];
376376
(bool, string, Parameter) triple = (false, "", param);
377377
if (data_names.ContainsKey(name))
@@ -405,10 +405,10 @@ private void BuildCache(NDArrayOrSymbolList args, bool update_graph = true)
405405
throw new Exception("A parameter was added to the graph during optimization but it was not added to the parameter dicts.\nPlease check the backend.");
406406
}
407407

408-
param = new Parameter(name, dtype: param_data.DataType);
408+
param = new Parameter(name, dtype: param_data.dtype);
409409
param._var_name = name;
410410
serialization_name = name;
411-
param.LoadInit(param_data, new Context[] { param_data.Context });
411+
param.LoadInit(param_data, new Context[] { param_data.ctx });
412412
}
413413

414414
triple = (false, serialization_name, param);
@@ -511,7 +511,7 @@ internal NDArrayOrSymbolList CallCachedOp(NDArrayOrSymbolList args)
511511
return Regroup(new List<NDArrayOrSymbol[]> { @out.NDArrayOrSymbols }, _out_format).Item1;
512512
}
513513

514-
public void OptimizeFor(NDArray x, string backend = null, bool clear = false, bool partition_if_dynamic = true, bool static_alloc = false,
514+
public void OptimizeFor(ndarray x, string backend = null, bool clear = false, bool partition_if_dynamic = true, bool static_alloc = false,
515515
bool static_shape = false, int inline_limit = 2, int? forward_bulk_size = null, int? backward_bulk_size = null, Dictionary<string, string> backend_opts = null, NDArrayList args = null)
516516
{
517517
this._backend = backend;
@@ -538,7 +538,7 @@ public void OptimizeFor(NDArray x, string backend = null, bool clear = false, bo
538538
var ctx_set = _tup_1.Item3;
539539
if (!has_symbol && !has_ndarray)
540540
{
541-
throw new Exception("In HybridBlock, there must be one NDArray or one Symbol in the input. Please check the type of the args.\n");
541+
throw new Exception("In HybridBlock, there must be one ndarray or one _Symbol in the input. Please check the type of the args.\n");
542542
}
543543
if (ctx_set.Length > 1)
544544
{
@@ -745,14 +745,14 @@ public void InferType(NDArrayOrSymbolList args)
745745
var params_filename = String.Format("%s-%04d.params", path != null ? path : "", epoch);
746746
if (path != null)
747747
{
748-
NDArray.Save(params_filename, arg_dict);
748+
ndarray.Save(params_filename, arg_dict);
749749
return (sym_filename, arg_dict.Count > 0 ? params_filename : null);
750750
}
751751

752752
return ("", "");
753753
}
754754

755-
public (Symbol, NDArrayDict) Export(int epoch = 0, bool remove_amp_cast = true)
755+
public (_Symbol, NDArrayDict) Export(int epoch = 0, bool remove_amp_cast = true)
756756
{
757757
if (_cached_graph == null)
758758
throw new Exception("Please first call block.hybridize() and then run forward with " +
@@ -836,7 +836,7 @@ public override NDArrayOrSymbol Forward(NDArrayOrSymbol x, params NDArrayOrSymbo
836836

837837
if (!has_symbol && !has_ndarray)
838838
{
839-
throw new Exception("In HybridBlock, there must be one NDArray or one Symbol in the input. Please check the type of the args.\n");
839+
throw new Exception("In HybridBlock, there must be one ndarray or one _Symbol in the input. Please check the type of the args.\n");
840840
}
841841

842842
if (has_ndarray)

‎src/MxNet/Gluon/Constant.cs‎

Lines changed: 6 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -15,12 +15,13 @@ limitations under the License.
1515
******************************************************************************/
1616
using System;
1717
using MxNet.Initializers;
18+
using MxNet.Numpy;
1819

1920
namespace MxNet.Gluon
2021
{
2122
public class Constant : Parameter
2223
{
23-
public Constant(string name, NDArray value) : base(name, OpGradReq.Null, value.Shape, value.DataType,
24+
public Constant(string name, ndarray value) : base(name, OpGradReq.Null, value.shape, value.dtype,
2425
init: new CInit(value))
2526
{
2627
Value = value;
@@ -37,20 +38,20 @@ public override OpGradReq GradReg
3738
}
3839
}
3940

40-
public NDArray Value { get; set; }
41+
public ndarray Value { get; set; }
4142

4243
public string InitName { get; set; }
4344

4445
public class CInit : Initializer
4546
{
46-
private readonly NDArray _value;
47+
private readonly ndarray _value;
4748

48-
public CInit(NDArray value)
49+
public CInit(ndarray value)
4950
{
5051
_value = value;
5152
}
5253

53-
public override void InitWeight(string name, ref NDArray arr)
54+
public override void InitWeight(string name, ref ndarray arr)
5455
{
5556
_value.CopyTo(arr);
5657
}

‎src/MxNet/Gluon/Contrib/Data/Vision/Dataloader/DataLoader.cs‎

Lines changed: 3 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,6 @@
11
using MxNet.Gluon.NN;
22
using MxNet.Image;
3+
using MxNet.Numpy;
34
using System;
45
using System.Collections.Generic;
56
using System.Text;
@@ -9,14 +10,14 @@ namespace MxNet.Gluon.Contrib.Data.Vision
910
public class DataLoader
1011
{
1112
public static HybridSequential CreateImageAugment(Shape data_shape, int resize= 0, bool rand_crop= false, bool rand_resize= false,
12-
bool rand_mirror= false, NDArray mean= null, NDArray std= null, float brightness= 0, float contrast= 0, float saturation= 0, float hue= 0,
13+
bool rand_mirror= false, ndarray mean= null, ndarray std= null, float brightness= 0, float contrast= 0, float saturation= 0, float hue= 0,
1314
float pca_noise= 0, float rand_gray= 0, int inter_method= 2, DType dtype= null)
1415
{
1516
throw new NotImplementedRelease1Exception();
1617
}
1718

1819
public static Sequential CreateBboxAugment(Shape data_shape, float rand_crop= 0, float rand_pad= 0, float rand_gray= 0,
19-
bool rand_mirror= false, NDArray mean= null, NDArray std= null, float brightness= 0, float contrast= 0,
20+
bool rand_mirror= false, ndarray mean= null, ndarray std= null, float brightness= 0, float contrast= 0,
2021
float saturation= 0, float pca_noise= 0, float hue= 0, int inter_method= 2,
2122
float max_aspect_ratio= 2, (float, float)? area_range= null,
2223
int max_attempts= 50, (int, int, int)? pad_val= null, DType dtype = null)

‎src/MxNet/Gluon/Contrib/Data/Vision/Dataloader/ImageBboxDataLoader.cs‎

Lines changed: 4 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -1,26 +1,27 @@
11
using MxNet.Gluon.Data;
22
using MxNet.Image;
3+
using MxNet.Numpy;
34
using System;
45
using System.Collections;
56
using System.Collections.Generic;
67
using System.Text;
78

89
namespace MxNet.Gluon.Contrib.Data.Vision
910
{
10-
public class ImageBboxDataLoader : IEnumerable<(NDArray, NDArray)>
11+
public class ImageBboxDataLoader : IEnumerable<(ndarray, ndarray)>
1112
{
1213
public int Length => throw new NotImplementedRelease1Exception();
1314

1415
public ImageBboxDataLoader(int batch_size, Shape data_shape, string path_imgrec= null, string path_imglist= null, string path_root= ".",
1516
int part_index= 0, int num_parts= 1, Augmenter[] aug_list= null, List<(string, float[])> imglist= null, bool coord_normalized = false,
1617
DType dtype= null, bool shuffle= false, Sampler<int> sampler= null, string last_batch= null, BatchSampler batch_sampler= null,
17-
Func<(NDArray, NDArray)[], (NDArray, NDArray)> batchify_fn = null, int num_workers= 0, bool pin_memory= false,
18+
Func<(ndarray, ndarray)[], (ndarray, ndarray)> batchify_fn = null, int num_workers= 0, bool pin_memory= false,
1819
int pin_device_id= 0, int? prefetch= null, bool thread_pool= false, int timeout= 120)
1920
{
2021
throw new NotImplementedRelease1Exception();
2122
}
2223

23-
public IEnumerator<(NDArray, NDArray)> GetEnumerator()
24+
public IEnumerator<(ndarray, ndarray)> GetEnumerator()
2425
{
2526
throw new NotImplementedRelease1Exception();
2627
}

‎src/MxNet/Gluon/Contrib/Data/Vision/Dataloader/ImageDataLoader.cs‎

Lines changed: 4 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -1,26 +1,27 @@
11
using MxNet.Gluon.Data;
22
using MxNet.Image;
3+
using MxNet.Numpy;
34
using System;
45
using System.Collections;
56
using System.Collections.Generic;
67
using System.Text;
78

89
namespace MxNet.Gluon.Contrib.Data.Vision
910
{
10-
public class ImageDataLoader : IEnumerable<(NDArray, NDArray)>
11+
public class ImageDataLoader : IEnumerable<(ndarray, ndarray)>
1112
{
1213
public int Length => throw new NotImplementedRelease1Exception();
1314

1415
public ImageDataLoader(int batch_size, Shape data_shape, string path_imgrec= null, string path_imglist= null, string path_root= ".",
1516
int part_index= 0, int num_parts= 1, Augmenter[] aug_list= null, List<(string, float[])> imglist= null,
1617
DType dtype= null, bool shuffle= false, Sampler<int> sampler= null, string last_batch= null, BatchSampler batch_sampler= null,
17-
Func<(NDArray, NDArray)[], (NDArray, NDArray)> batchify_fn = null, int num_workers= 0, bool pin_memory= false,
18+
Func<(ndarray, ndarray)[], (ndarray, ndarray)> batchify_fn = null, int num_workers= 0, bool pin_memory= false,
1819
int pin_device_id= 0, int? prefetch= null, bool thread_pool= false, int timeout= 120)
1920
{
2021
throw new NotImplementedRelease1Exception();
2122
}
2223

23-
public IEnumerator<(NDArray, NDArray)> GetEnumerator()
24+
public IEnumerator<(ndarray, ndarray)> GetEnumerator()
2425
{
2526
throw new NotImplementedRelease1Exception();
2627
}

0 commit comments

Comments
 (0)