MSP430FR5994: Implementing a neural network on MSP430FR5994

I am trying to implement a neural network model on a MSP430FR5994 - the model will be pre-trained offline and only needs to run on the MCU. The purpose of the project is to capture audio signals - process them through FFT and then once they are passed through the model - the resultant code needs to be uploaded back to the MCU. 

I have already managed to convert some sample keras models into C files which contain the weights and biases of the neural network used to process the signals. Since this is my first time working with the MSP430, I am not exactly sure how these C files which contain the weights can be uploaded onto the MCU. One method that I was considering included a converter called - keras2c - which processes the entire model and then generates a complete C executable file - the resultant header and C files could be loaded into the MSP as a library which could be used for the project. 

While seaerching across the forum for similar questions, I found a thread that mentions uploading the pre-trained model into the FRAM of the MCU - however, I am not completely clear about this process and would highly appreciate if I could get a bit more details or clarity about this method.  

If there is any more information needed to clarify the understanding of the issue, please feel free to write so in the replies as this is my first question on the forum - I would gladly update the thread accordingly.