rdturnermtl

Currently, `pipreqs` uses the host system when compiling requirements (OS and Python version). There should be options to compile what the requirements would be on any specific Python version and...

The mappings file could use the entry ``` pyro:pyro-ppl ``` The `pipreqs` tool currently does not work with `pyro` see [pyro issue #2018](https://github.com/pyro-ppl/pyro/issues/2018).

Put the import of `gitpython` in a try-except because it is not needed for the version number when installing off the wheel (`PyPI`).

Allow the user to call experiments with an argument file: commands can be called like `bayesmark-exp --args args.yaml` instead of specifying arguments. All the arguments are simply in the `yaml`...

Allow the user to use `yaml` instead of `json`, if they prefer, for the `config.json` file.

It would be nice if `scspell` supported the pre-commit tool as has been done with [codespell](https://github.com/codespell-project/codespell). We just need to add a `.pre-commit-hooks.yaml` file at the top level. It would...

Current report shows 70%, that number should be higher.

The nominal coverage of precision recall curves doesn't match the 95% specified level in boot_test.py. We must set ``x_grid = np.linspace(0.05, 0.95, DEFAULT_NGRID)`` to get the tests to pass. We...

There are already lots of randomized inputs in the tests, this can be made cleaner using the hypothesis package: [http://hypothesis.readthedocs.io/en/latest/index.html](url) We should also make an effort to randomize input types,...

Add the loss function for accuracy-at-k where multiple predictions are provided. The correctness of the decision rule can be tested against `classification.hard_loss` with the appropriate (large) loss matrix.