Psoc ML loading issues

[IulianMurariu-T]

Hello. I have a Psoc 6 CY8CKIT-062S2-43012 and I'm trying to load a keras model. I am having trouble loading a keras model with a custom layer into our application. The ML Configurator returns us an error: "Error returned: Unknown layer: FFT_Image. Please ensure this object is passed to the custom_objects argument."

Even when we don't use a custom layer, we still get the error: "Deploy script returned with error 'list' object has no attribute 'name'".

And finally in the project itself , in the library file "mtb_ml_common.h", we get a linker error: " #error Unsupported data type" We've been looking everywhere in the exemple projects and cannot figure out where COMPONENT_ML_FLOAT32 is defined and how can we define it properly in our project as well. We would really appreciate some help.

[RodolfoGL · Accepted solution]

Hello. At the moment, we don't support this type of layer: "FFT_image". 

This is a list of layers we support:

                               'Add',
                                'AveragePooling1D',
                                'AveragePooling2D',
                                'BatchNormalization',
                                'Concatenate',
                                'Conv1D',
                                'Conv2D',
                                'Dense',
                                'DepthwiseConv2D',
                                'Dropout',
                                'Flatten',
                                'GlobalAveragePooling1D',
                                'GlobalAveragePooling2D',
                                'GlobalMaxPooling1D',
                                'GlobalMaxPooling2D',
                                'GRU',
                                'InputLayer',
                                'LeakyReLU',
                                'LSTM',
                                'MaxPooling1D',
                                'MaxPooling2D',
                                'ReLU',
                                'Reshape',
                                'SeparableConv2D',
                                'Softmax']

If you refer the mtb-example-profiler code example Makefile, we add the COMPONENT_ML_FLOAT32 in here:

ifeq (float, $(NN_TYPE))

COMPONENTS+=ML_FLOAT32

endif