Relax size constraint when creating a tensor from an array.

Author: NiklasGustafsson

src/TorchSharp/Tensor/Factories/Tensor.Factories.cs

@@ -143,7 +143,7 @@ private static Tensor _tensor_generic(Array rawArray, ReadOnlySpan<long> dimensi
                 if (origType == (sbyte)ScalarType.ComplexFloat32)
                     prod *= 2;
 
-                if (prod != rawArray.LongLength)
+                if (prod > rawArray.LongLength)
                     throw new ArgumentException($"mismatched total size creating a tensor from an array: {prod} vs. {rawArray.LongLength}");
             }
 

test/TorchSharpTest/TestTorchTensor.cs

@@ -783,6 +783,16 @@ public void FromArrayFactory()
                         () => Assert.Equal(ScalarType.Bool, t.dtype));
                 }
 
+                {
+                    var array = new bool[18]; // Too long, on purpose
+                    var t = torch.tensor(array, new long[] { 8 }, device: device);
+                    Assert.Multiple(
+                        () => Assert.Equal(8, t.NumberOfElements),
+                        () => Assert.Equal(device.type, t.device_type),
+                        () => Assert.Equal(1, t.ndim),
+                        () => Assert.Equal(ScalarType.Bool, t.dtype));
+                }
+
                 {
                     var array = new int[8];
                     var t = torch.tensor(array, device: device);
@@ -801,6 +811,16 @@ public void FromArrayFactory()
                         () => Assert.Equal(ScalarType.Float32, t.dtype));
                 }
 
+                {
+                    var array = new float[18]; // Too long, on purpose
+                    var t = torch.tensor(array, new long[] { 8 }, device: device);
+                    Assert.Multiple(
+                        () => Assert.Equal(8, t.NumberOfElements),
+                        () => Assert.Equal(device.type, t.device_type),
+                        () => Assert.Equal(1, t.ndim),
+                        () => Assert.Equal(ScalarType.Float32, t.dtype));
+                }
+
                 {
                     var array = new double[1, 2];
                     var t = torch.from_array(array, device: device);