Scvi

------------------

* corrected notebook `jules`_
* added UMAP and updated harmonization code `chenling`_ `romain`_
* support for batch indices in csvdataset `primoz-godec`_
* speeding up likelihood computations `william-yang`_
* better anndata interop `casey-greene`_
* early stopping based on classifier accuracy `david-kelley`_

------------------

* updated to torch v1 `jules`_
* added stress tests for harmonization `chenling`_
* fixed autograd breaking `romain`_
* make removal of empty cells more efficient `john-reid`_
* switch to os.path.join `casey-greene`_

------------------

* added baselines and datasets for sMFISH imputation `jules`_
* added harmonization content `chenling`_
* fixing bugs on DE `romain`_

------------------

* annotation notebook `eddie`_
* Memory footprint management `jeff`_
* updated early stopping `max`_
* docstring `james-webber`_

------------------

* MMD and adversarial inference wrapper `eddie`_
* Documentation `jeff`_
* smFISH data imputation `max`_

------------------

* Dataset additions `eddie`_
* Documentation `yining`_
* updated early stopping `max`_