Jeff Kinnison

Installing from PyPI does not create a .shadhorc file in $HOME. This should be created automatically on first use of SHADHO if it does not exist.

Create an installed executable to install/uninstall Work Queue and the .shadho directory.

Currently, SHADHO requires a lot of manual configuration. Much of this can be pushed behind-the-scenes (e.g. Work Queue config) or be automated (e.g. a worker type function for running the...

See title. What are you doing with your life, Jeff???

See title. Need documentation for all of this.

See title. Need good docs up and running.

[WIP] Create a binary decoder with a TART reasoning module

3

This adds a TART reasoning module as a binary decoder. TODOs: [] Add the vanilla and LOO embedding protocols [] Add an embeddings cache [] Tests

fix: Count full dataset in DatasetShard.__len__

1

Running the new test with the previous code, `DatasetShard.__len__` returned `256` rather than the full dataset size defined in the test. Summing the size of the batches returned by `DatasetIterator.iter_batches`...

fix: Copy modules and alias weights data when using tied weights

3

The current tied weights implementation aliases torch modules between tied input features. Best practices seem to be deep copying the modules and then aliasing the underlying weights data (`Parameter.data`) between...