… multi-output training loops.

… our new Molecule data structure.

…ength.

…and that's all.

…nly difference is the batching function, which comes from malt.data.utils.

…h feature extractors.

Adding Random and Thompson Sampling Bayes Opt Policies

…d data structures.

Data with tests. Need to merge change to collections, though.

shouldn't live on github

Only made the pin_mem False to debug something.

Updating the collate function

…nd data APIs.

…ending to Dataset(). It somehow sets a new class init signature and every time after that if you call Dataset(), it'll give you the updated list, instead of a fresh [].

…data. Also, fixed the training loop and benchmarked using some new notebooks. Also added multi-objective stepping.

refactor the gpytorch implementation

GPyTorch