NumSharp.Backends.Unmanaged.UnmanagedMemoryBlock`1 fails on Mono on Linux

Mono 5.12.0.301
Linux 4.19.76

Attempting the "Hello World" Tensortflow.NET example in a docker environment running Mono yields the following error. Exactly the same code on Mono 6.12.0.93 on Mac OS X however does run flawlessly

2020-11-15 17:53:41.843832: I tensorflow/core/platform/cpu_feature_guard.cc:142] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN)to use the following CPU instructions in performance-critical operations: AVX2 FMA
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
2020-11-15 17:53:41.873442: I tensorflow/core/platform/profile_utils/cpu_utils.cc:104] CPU Frequency: 2712000000 Hz
2020-11-15 17:53:41.874170: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x7fc874d557a0 initialized for platform Host (this does not guarantee that XLA will be used). Devices:
2020-11-15 17:53:41.874230: I tensorflow/compiler/xla/service/service.cc:176] StreamExecutor device (0): Host, Default Version
Stacktrace:

at <0xffffffff>

• Assertion at method-to-ir.c:7352, condition `!sig->has_type_parameters' not met

  at NumSharp.Backends.Unmanaged.UnmanagedMemoryBlock`1.FromArray (byte[],bool) [0x00037] in <6d1fbec37f814ff9b52dec21dc0ebd1a>:0
  at NumSharp.Backends.Unmanaged.ArraySlice.FromArray (byte[],bool) [0x00000] in <6d1fbec37f814ff9b52dec21dc0ebd1a>:0
  at NumSharp.np.array (byte[]) [0x00000] in <6d1fbec37f814ff9b52dec21dc0ebd1a>:0
  at Tensorflow.Tensor.GetNDArray (Tensorflow.TF_DataType) [0x00041] in :0
  at Tensorflow.Tensor.numpy () [0x00007] in :0
  at Tensorflow.tensor_util.to_numpy_string (Tensorflow.Tensor) [0x00034] in :0
  at Tensorflow.Eager.EagerTensor.ToString () [0x00016] in :0

Start with the stack trace at NumSharp.Backends.Unmanaged.UnmanagedMemoryBlock.FromArray and follow the calls through ArraySlice.FromArray, np.array, Tensor.GetNDArray, Tensor.numpy, and EagerTensor.ToString. Reproduce the Hello World TensorFlow.NET example in the reported Mono Linux Docker environment and compare it with the working Mono macOS case. Done means the Mono assertion no longer occurs and the example completes successfully.