r""" Build a batch of latent vectors for the generator. Args: n_samples (int): number of vector in the batch

r""" Load a model saved with the @method save() function Args: - path (string): file where the model is stored

r""" Retrieve the original G network. Use this function when you want to modify G after the initialization

r""" Generate some data given the input latent vector. Args: input (torch.tensor): input latent vector

r""" Retrieve the original D network. Use this function when you want to modify D after the initialization

r""" Compute the classification penalty associated with the current output Args: - outputD (tensor): discriminato

r""" Parse a file name with the given pattern: pattern = ($model_name)_s($scale)_i($iteration).pt Returns: None if the path doesn't fulfi

r""" Load a model saved with the @method save() function Args: - in_state (dict): state dict containing the model

r""" Update the value of the merging factor alpha Args: - alpha (float): merging factor, must be in [0, 1]

r""" Load the input state dict to the module except for the weights corresponding to one of the forbidden layers

r""" Save the model at the given location. All parameters included in the self.config class will be saved as well. Args:

r""" Move the current networks and solvers to the GPU. This function must be called each time netG or netD is modified

r""" Gradient penalty as described in "Improved Training of Wasserstein GANs" https://arxiv.org/pdf/1704.00028.pdf Args: - i

r""" Add a new scale to the model. Increasing the output resolution by a factor 2 Args: - depthNewScale (int): de

r""" Create and upload a moving average generator.

r""" Size of the latent vector given by self.buildRandomCriterionTensor

r""" If the dataset is labelled, give the order in which the attributes are given Args: - shiftAttrib (dict): order of each category

r""" If the dataset is labelled, give the order in which the attributes are given Returns: A dictionary output[k

r""" Get the last checkpoint of the model with name @param name detected in the directory (@param dir) Returns: trainConfig, pathMode

r""" The discrimator should be defined here.

r""" The generator should be defined here.

r""" The format layer represents the first weights applied to the latent vector. It converts a 1xdimLatent input into a 4 x 4 x scales

r""" Initialize the GAN model.

r""" For labelled datasets: initialize the classification criterion.

r""" Load the input state dict to the module except for the weights corresponding to one of the forbidden layers

r""" Update the discrimator D using the given "real" inputs. Args: input (torch.tensor): input batch of real data

r""" Load a permanent configuration describing a models. The variables described in this file are constant through the training.

r""" Launch the training. This one will stop if a divergent behavior is detected. Returns: - True if the trainin

r""" Given the number of iterations between two updates of alpha at each scale and the number of updates per scale, build the effectiv

r""" Update a configuration with the fields of another given configuration

Creates a data manager @arg

r""" Build a generator for a progressive GAN model Args: - dimLatent (int): dimension of the latent vector -

r""" If a starting scale is defined other than 0, this method adds to the network the first 0-startingScale blocks.

r""" Build a batch of vectors with a random combination of the values of the existing classes Args: sizeBatch (in

r""" Using a new configuration dictionary and a default configuration setup an object with the given configuration. for example, if you h

r""" Load the training dataset for the given scale. Args: - scale (int): scale at which we are working Returns:

r""" Retrieve all the members of obj which are also members of referenceConfig and dump them into a dictionnary If printDefault is activa

r""" Get He's constant for the given layer https://www.cv-foundation.org/openaccess/content_iccv_2015/papers/He_Delving_Deep_into_ICCV_2015_pa

r""" Optimizer of the discriminator.

r""" Optimizer of the generator.

r""" Get the size of the generated image.

r""" Get the model's parameters