ONNX ORT_INVALID_ARGUMENT(2) exception when creating tensor with Ort::Value::CreateTensor

Question

I am trying to create a tensor for ONNX with Ort::Value::CreateTensor using the following snippet:

Ort::MemoryInfo MemoryInfo = Ort::MemoryInfo::CreateCpu(OrtAllocatorType::OrtDeviceAllocator, OrtMemType::OrtMemTypeCPU);


std::vector<Ort::Value> vInputTensors;
    
for (std::vector<float> vfInput : TestDataset.GetInputs()) {
    unsigned uSequenceLength = vfInput.size();

    // Shape: (batch_size=1, sequence_length, num_norm_attributes, num_channels=1)
    const std::vector<int64_t> vi64InputShape = {BATCH_SIZE, uSequenceLength, 2 * NUM_ATTRIBUTES, NUM_CHANNELS};

    Ort::Value::CreateTensor<float>(MemoryInfo, vfInput.data(), uSequenceLength, vi64InputShape.data(), vi64InputShape.size());
    //vInputTensors.push_back(Ort::Value::CreateTensor<float>(MemoryInfo, vfInput.data(), uSequenceLength, vi64InputShape.data(), vi64InputShape.size()));
}

I have printed the elements in vfInput.data() and vi64InputShape.data() and printed the values of uSequenceLength and vi64InputShape.size(). And everything is a expected.

However, the program throws an exception:

Unhandled exception at 0x00007FFC7ABDCF19 in TestApp.exe: Microsoft C++ exception: Ort::Exception at memory location 0x00000036901AF6B0.

When I change BATCH_SIZE, uSequenceLength, 2 * NUM_ATTRIBUTES, NUM_CHANNELS} to {BATCH_SIZE, 100, 2 * NUM_ATTRIBUTES, NUM_CHANNELS}, it doesn't complain but of course that's not correct.

Please may you assist me with this.

Thank you.