Keras subclass-based model CAN NOT save model/weights

[quanap5kr]

Hi,

I am interested in Keras Api using the Subclass style for building the model. I clone example and run example MNIST CNN (Keras Subclass).
I use save_weights function but get errors as below.
Could you give any idea to save the model or weights? Thank you very much.

• Snip of code:

            {
                x_test = x_test["::100"];
                y_test = y_test["::100"];
                var pred = conv_net.Apply(x_test);
                accuracy_test = (float)accuracy(pred, y_test);
                print($"Test Accuracy: {accuracy_test}");
            }

            conv_net.save_weights("subclass_mnist_model"); // I added this
           
            return accuracy_test > 0.90;

• Error throw:

Choose one example to run, hit [Enter] to run all: 15
2021-09-08 오전 8:12:50 Starting MNIST CNN (Keras Subclass)
The file C:\Users\Quan\AppData\Local\Temp\mnist.npz already exists
2021-09-08 17:12:51.202274: 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
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
step: 10, loss: 1.7204771, accuracy: 0.83
step: 20, loss: 1.5916963, accuracy: 0.898
step: 30, loss: 1.5561028, accuracy: 0.932
step: 40, loss: 1.5472074, accuracy: 0.944
step: 50, loss: 1.5292175, accuracy: 0.963
step: 60, loss: 1.507012, accuracy: 0.978
step: 70, loss: 1.5141329, accuracy: 0.968
step: 80, loss: 1.4908792, accuracy: 0.989
step: 90, loss: 1.4994849, accuracy: 0.978
Test Accuracy: 0.96
HDF5-DIAG: Error detected in HDF5 (1.10.5) thread 37204:
  #000: C:\projects\hdf-pinvoke-1-10\CMake-hdf5-1.10.5\hdf5-1.10.5\src\H5L.c line 790 in H5Lexists(): not a location
    major: Invalid arguments to routine
    minor: Inappropriate type
  #001: C:\projects\hdf-pinvoke-1-10\CMake-hdf5-1.10.5\hdf5-1.10.5\src\H5Gloc.c line 246 in H5G_loc(): invalid object ID    major: Invalid arguments to routine
    minor: Bad value
HDF5-DIAG: Error detected in HDF5 (1.10.5) thread 37204:
  #000: C:\projects\hdf-pinvoke-1-10\CMake-hdf5-1.10.5\hdf5-1.10.5\src\H5D.c line 119 in H5Dcreate2(): not a location ID    major: Invalid arguments to routine
    minor: Inappropriate type
  #001: C:\projects\hdf-pinvoke-1-10\CMake-hdf5-1.10.5\hdf5-1.10.5\src\H5Gloc.c line 246 in H5G_loc(): invalid object ID    major: Invalid arguments to routine
    minor: Bad value
HDF5CSharp.DataTypes.Hdf5Exception: Unable to create dataset for kernel:0
   at HDF5CSharp.Hdf5Utils.GetId(Int64 parentId, String name, Int64 dataType, Int64 spaceId, Hdf5ElementType type)
   at HDF5CSharp.Hdf5Utils.GetDatasetId(Int64 parentId, String name, Int64 dataType, Int64 spaceId)
   at HDF5CSharp.Hdf5.WriteDatasetFromArray[T](Int64 groupId, String name, Array dset)
   at Tensorflow.Keras.Saving.hdf5_format.WriteDataset(Int64 f, String name, Tensor data)
   at Tensorflow.Keras.Saving.hdf5_format.save_weights_to_hdf5_group(Int64 f, List`1 layers)
   at Tensorflow.Keras.Engine.Model.save_weights(String filepath, Boolean overwrite, String save_format, Object options)   at TensorFlowNET.Examples.DigitRecognitionCnnKeras.Run() in E:\Tensorflow.Net\Keras.NetExample\src\TensorFlowNET.Examples\ImageProcessing\MnistCnnKerasSubclass.cs:line 97
   at TensorFlowNET.Examples.Program.RunExamples(String key, IExample[] examples) in E:\Tensorflow.Net\Keras.NetExample\src\TensorFlowNET.Examples\Program.cs:line 89
2021-09-08 오전 8:13:20 Completed MNIST CNN (Keras Subclass) ++
Example: MNIST CNN (Keras Subclass) is Failed!
TensorFlow Binary v2.6.0
TensorFlow.NET v0.60.2.0
TensorFlow.Keras v0.6.2.0
1 of 27 example(s) are completed.
Press [Enter] to continue...