SVGP Model Updating how to merge fantasy_model parameters to the original SVGP model

[Seas00n]

Hi, I see https://docs.gpytorch.ai/en/stable/examples/08_Advanced_Usage/SVGP_Model_Updating.html
And I have a question if I want to use the third batch data to train the GP model, how should I merge the parameters of cond_model and the original model such that I can get a new model and generate a new cond_model that doesn't forget all data before and fit the new add third batches data.

val_x = torch.linspace(3, 5, 25).view(-1,1)
val_y = torch.sin(6. * val_x) + 0.3 * torch.randn_like(val_x)
cond_model = model.variational_strategy.get_fantasy_model(inputs=val_x, targets=val_y.squeeze())
# ......
# third batch data
val_x2 = torch.linspace(5, 8, 25).view(-1,1)
val_y2 = torch.sin(6. * val_x) + 0.3 * torch.randn_like(val_x)
# how to merge cond_model and model and generate a new cond_model
cond_model2 = ?