Ubuntu Tensor indexer null exception

[yorkshireandrew]

Description

My system is:

Unix 5.15.0.88 64Bit Operating System: True .NET CLR: 6.0.25 TensorFlow Binary v2.15.0-rc1 TensorFlow.NET v0.150.0.0 TensorFlow.Keras v0.15.0.0 Ubuntu 20.04.6 LTS

My code amounts to this:

using Tensorflow;
using Tensorflow.Keras.Metrics;
using static Tensorflow.Binding;

namespace TensorFlowNET.Examples
{
    class Program
    {
        static void Main(string[] args){
            var gtArray = new float[4]{1.0F, 3.0F, 1.0F, 3.0F};
            var gtTensor = new Tensor(gtArray);
            var a = gtTensor.ToArray<float>();
            var h1 = gtTensor[1];
        }
    }
}

While a produces the expected float array, h1 produces the exception below.

Exception has occurred: CLR/System.NullReferenceException An unhandled exception of type 'System.NullReferenceException' occurred in Tensorflow.Binding.dll: 'Object reference not set to an instance of an object.' at Tensorflow.Operation.op_Implicit(Operation op) at Tensorflow.Tensor._as_tf_output() at Tensorflow.ops._create_c_op(Graph graph, NodeDef node_def, Tensor[] inputs, Operation[] control_inputs, OpDef op_def) at Tensorflow.Operation..ctor(NodeDef node_def, Graph g, Tensor[] inputs, TF_DataType[] output_types, ITensorOrOperation[] control_inputs, TF_DataType[] input_types, String original_op, OpDef op_def) at Tensorflow.Graph.create_op(String op_type, Tensor[] inputs, TF_DataType[] dtypes, TF_DataType[] input_types, String name, Dictionary2 attrs, OpDef op_def, Boolean compute_device) at Tensorflow.OpDefLibrary._apply_op_helper(String op_type_name, String name, Dictionary2 keywords) at Tensorflow.Contexts.Context.ExecGraphAction(String OpType, String Name, ExecuteOpArgs args) at Tensorflow.Contexts.Context.ExecuteOp(String opType, String name, ExecuteOpArgs args) at Tensorflow.array_ops.stridedslice(Tensor input, Tensor begin, Tensor end, Tensor strides, Int32 begin_mask, Int32 end_mask, Int32 ellipsis_mask, Int32 new_axis_mask, Int32 shrink_axis_mask, String name) at Tensorflow.Tensor.<>c__DisplayClass146_0.b__0(NameScope scope) at Tensorflow.Binding.tf_with[TIn,TOut](TIn py, Func`2 action) at Tensorflow.Tensor.slice(Int32 start) at Tensorflow.Tensor.get_Item(Int32 idx)

Am I missing part of what is required to make TensorFlow.NET work. This for example gives me no results, but I don't think it is required is it? pip show tensorflow

Reproduction Steps

No response

Known Workarounds

No response

Configuration and Other Information

No response

[yorkshireandrew]

Here is a breaking test file seems to error for Release v0.150.0 based on tensorflowv v2.15.0.

using Microsoft.VisualStudio.TestTools.UnitTesting;
using Tensorflow.NumPy;
using System.Linq;
using Tensorflow;
using static Tensorflow.Binding;

namespace TensorFlowNET.UnitTest.Basics
{   
    [TestClass]
    public class TensorTest
    {
         [TestMethod]
        public void TensorIndexingTest()
        {
            var elements = new int[3]{1, 2 , 3};
            var tensor = new Tensor(elements);
            var firstElement = (int) tensor[0];
            var secondElement = (int) tensor[1];
                        Assert.AreEqual(1, firstElement);
            Assert.AreEqual(2, secondElement);
        }
    }
}

[Wanglongzhi2001]

Hello, please create tensor from NDArray in your case. For more information about how to create a tensor, you can refer to the source code. https://github.com/SciSharp/TensorFlow.NET/blob/master/src/TensorFlowNET.Core/Tensors/Tensor.Creation.cs

[yorkshireandrew]

I was writing unit tests around a method that takes a 1D Tensor then extracts each element from it (each one still as a Tensor) I think the code above fails because the Tensor has not yet been calculated in a graph.

If I make this replacement in the test above it then passes:

var tensor = tf.constant(elements);

[yorkshireandrew]

@Wanglongzhi2001 your right constructing the Tensor from NDArray also works fine, thankyou.

    var raw_elements = new int[3]{1, 2 , 3};
        var elements = np.array(raw_elements);
        var tensor = new Tensor(elements);

I'll close the issue. What a great project. I'm finding using BARD rather than ChatGPT to help with coding works better as the code examples it generates are more up to date.