I'm currently delving into the workings of a library hosted on a Git repository, specifically focusing on its use of GPFlow for model initialization. The initialization code snippet is as follows:
gpflow.models.GPR(data=(training_params.values, wrapped_data.data), kernel=kernel, noise_variance=tf.constant(noise_variance, type=wrapped_data.dtype), **kwargs)
I've encountered issues with 'wrapped_data.data', particularly because it's a masked array, and I'm unsure of GPFlow's approach to processing this type of data structure.
I'm expecting that GPFlow is handling these data as a regular numpy array, without a mask.