update resample in DirectMLX (#495)

Libraries/DirectMLX.h

@@ -3322,6 +3322,7 @@ namespace dml
         Expression input,
         TensorDimensions outputSizes,
         DML_INTERPOLATION_MODE mode,
+        DML_AXIS_DIRECTION roundingDirection,
         Span<const float> scales = {},
         Span<const float> inputPixelOffsets = {},
         Span<const float> outputPixelOffsets = {})
@@ -3331,6 +3332,9 @@ namespace dml
         TensorDesc inputTensor = input.Impl()->GetOutputDesc();
         uint32_t dimensionCount = static_cast<uint32_t>(inputTensor.sizes.size());
         assert(outputSizes.size() == dimensionCount);
+        assert(scales.empty() || scales.size() == dimensionCount);
+        assert(inputPixelOffsets.empty() || inputPixelOffsets.size() == dimensionCount);
+        assert(outputPixelOffsets.empty() || outputPixelOffsets.size() == dimensionCount);
 
         SmallVector<float, 4> defaultScales;
         if (scales.empty())
@@ -3358,17 +3362,18 @@ namespace dml
 
         TensorDesc outputTensor(inputTensor.dataType, std::move(outputSizes), builder->GetTensorPolicy());
 
-        DML_RESAMPLE1_OPERATOR_DESC desc = {};
+        DML_RESAMPLE2_OPERATOR_DESC desc = {};
         desc.InputTensor = inputTensor.AsPtr<DML_TENSOR_DESC>();
         desc.OutputTensor = outputTensor.AsPtr<DML_TENSOR_DESC>();
         desc.InterpolationMode = mode;
-        desc.DimensionCount = static_cast<UINT>(scales.size());
+        desc.RoundingDirection = roundingDirection;
+        desc.DimensionCount = dimensionCount;
         desc.Scales = scales.data();
         desc.InputPixelOffsets = inputPixelOffsets.data();
         desc.OutputPixelOffsets = outputPixelOffsets.data();
 
         detail::NodeOutput* const inputs[] = { input.Impl() };
-        detail::NodeID node = builder->CreateOperatorNode(DML_OPERATOR_RESAMPLE1, &desc, inputs);
+        detail::NodeID node = builder->CreateOperatorNode(DML_OPERATOR_RESAMPLE2, &desc, inputs);
         detail::NodeOutput* output = builder->CreateNodeOutput(node, 0, std::move(outputTensor));
 
         return output;