Current Limitations of nn4mc

nn4mc Limitations:

• Models need to be Keras Sequential() models. We're currently working on parsing for Keras Functional API models and also ONNX models.
• Activations ought to be performed at the layer, not as a separate layer:

instead of:

...
model.add(Conv1D(....))
model.add(Activation('relu'))

do:

model.add(Conv1D(...., activation = 'relu'))

Currently Supported Layers Are:

• Dense
• Conv1D
• Conv2D
• MaxPooling1D
• MaxPooling2D
• GRU

Actively being developed:

• Keras Functional API Parser
• ONNX Parser
• simpleRNN
• LSTM