'GPR' object has no attribute 'X'

Question

I am trying to run the code from https://github.com/befelix/safe_learning/blob/master/examples/1d_example.ipynb . This code is written for gpflow version 0.4.0. I want to update this code for gpflow version 2.5.2. I changed the following line

gp = gpflow.gpr.GPR(np.empty((0, 2), dtype=safe_learning.config.np_dtype),
                    np.empty((0, 1), dtype=safe_learning.config.np_dtype),
                    kernel,
                    mean_function=mean_function)

to

gp = gpflow.models.GPR(data=(np.empty((0, 2), dtype=safe_learning.config.np_dtype),
                    np.empty((0, 1), dtype=safe_learning.config.np_dtype)),
                    kernel=kernel,
                    mean_function=mean_function)

then it creates a class object, dynamics = safe_learning.GaussianProcess(gp, name='gp_dynamics') which is defined as follows:

    """A GaussianProcess model based on gpflow.

    Parameters
    ----------
    gaussian_process : instance of gpflow.models.GPModel
        The Gaussian process model.
    beta : float
        The scaling factor for the standard deviation to create
        confidence intervals.

    
    """

    def __init__(self, gaussian_process, beta=2., name='gaussian_process'):
        """Initialization."""
        super(GaussianProcess, self).__init__(name=name)

        with tf.variable_scope(self.scope_name):
            self.n_dim = gaussian_process.X.shape[-1]
            self.gaussian_process = gaussian_process
            self.beta = float(beta)

            self.input_dim = gaussian_process.X.shape[1]
            self.output_dim = gaussian_process.Y.shape[1]

            self.hyperparameters = [tf.placeholder(config.dtype, [None])]
            self.gaussian_process.make_tf_array(self.hyperparameters[0])

            self.update_feed_dict()
...

when I run this class, I get the following error

dynamics = safe_learning.GaussianProcess(gp, name='gp_dynamics')
Traceback (most recent call last):

  Input In [39] in <cell line: 1>
    dynamics = safe_learning.GaussianProcess(gp, name='gp_dynamics')

  File D:\python\safe_learning-master\safe_learning-master\safe_learning\functions.py:490 in __init__
    self.n_dim = gaussian_process.X.shape[-1]

AttributeError: 'GPR' object has no attribute 'X'

Please kindly help me how this attribute has changed in new version (2.5.2) of gpflow. I am using the following link to upgrade https://gpflow.github.io/GPflow/develop/notebooks/gpflow2_upgrade_guide.html#Model-Class

Thanks in advance.

Jesper Nielsen · Accepted Answer · 2022-09-13 08:26:42Z

0

There's a big difference between 0.4.0 and 2.5.2. I don't think you should expect this to be an easy process.

model.X and model.Y has been replaced with model.data[0] and model.data[1] respectively. In fact model.data corresponds to the data parameter you passed to the GPR initialiser.

answered Sep 13, 2022 at 8:26

Jesper Nielsen

4703 silver badges5 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

valli Over a year ago

Thanks a lot. Changing to data[0], data[1] is working. I am getting couple of issues similar to above. For example, "'GPR' object has no attribute 'make_tf_array'". Can you suggest any link that list out the changes? Though it is not an easy process, it is a requirement for me to make it run. Thank you.

Jesper Nielsen Over a year ago

I'm afraid you're asking for change over too long time. You're asking for the difference between now and 5 years ago, and that covers two major versions. We had an upgrade guide from version 1.x.x to 2.x.x, but even that is deprecated. We also have a change log, but that only goes back to version 2.1.5. I'm sorry. I think you're going to have to dive into the source code and really understand GPflow, if you want to make this kind of update.

Collectives™ on Stack Overflow

'GPR' object has no attribute 'X'

1 Answer 1

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related