Jack Buckingham

[Bug] Posterior on fantasy models can be memory inefficient on batches

1

# 🐛 Bug When using a fantasy model with a batch dimension, and retrieving the posterior with a set using more batch dimensions, internally GPyTorch / PyTorch tries to create...

[Bug] `mean_cache` not attached with correct `args` in `get_fantasy_model`

2

# 🐛 Bug When a model is fantasized, the mean and covariance caches are recreated and reattached. However, the mean cache is attached with no `args`, while when it is...