d9069391dc7dfe7cf383815a439f478917eafa63,gpytorch/mlls/exact_marginal_log_likelihood.py,ExactMarginalLogLikelihood,forward,#ExactMarginalLogLikelihood#Any#Any#,30

Before Change

        super(ExactMarginalLogLikelihood, self).__init__(likelihood, model)

    def forward(self, output, target):
        if not isinstance(output.covar(), LazyVariable):
            output = GaussianRandomVariable(
                output.mean(), NonLazyVariable(output.covar())
            )
        mean, covar = self.likelihood(output).representation()
        n_data = target.size(-1)

        // Get log determininat and first part of quadratic form

After Change

        prior_mean, prior_covar = output.representation()
        prior_covar = prior_covar
        if not isinstance(prior_covar, LazyVariable):
            prior_covar = NonLazyVariable(prior_covar)

        output = GaussianRandomVariable(prior_mean, prior_covar)
        mean, covar = self.likelihood(output).representation()
        n_data = target.size(-1)

Instances