Is it possible to run this on a (recent) Mac, which does not support CUDA? I would have guessed setting --GPU 0 would not attempt to call CUDA, but it fails.

File "/Users/../Desktop/bopbtl/venv/lib/python3.7/site-packages/torch/cuda/__init__.py", line 61, in _check_driver
    raise AssertionError("Torch not compiled with CUDA enabled")
AssertionError: Torch not compiled with CUDA enable

Is your feature request related to a problem? Please describe.

By default k0s configures only cert based auth for API server and only the admin.conf with cluster admin role bound to it.

We should make available some subcommand to easily create a new cert and kubeconfig for a new user.

Describe the solution you'd like
k0s user create jussi --group foo,bar

That would dump ou

Describe the bug
The sidebar has overflow: scroll causing scrollbars always to be visible on Windows

To Reproduce
Steps to reproduce the behavior:

1. Go to documentation using Windows

Expected behavior
There should not be scrollbars before the content overflows

Screenshots
![image](https://user-images.githubusercontent.com/2255960/99193174-891e7c00-2777-11eb-8ceb-79

We've now gotten a fair number of "cheatsheet" entries in the books list. I'd like to propose that we move these to a separate page.

Reasons:

1. Will be easier to find cheatsheets
2. Cheatsheets aren't books
3. Cheat sheets aren't so spoken-language specific.
4. The english-language books page is getting long.
5. Will encourage better cheatsheet coverage.

Against:

1. Scope creep.

Errors:
126. new Intl.NumberFormat is not supposed to get passed a single object as param. The first param must be a string and the second param can be an object.
135. Spelling error ('tis' should be 'this').

Suggestion:
86. Needs to be more explicit by calling the function in the example. Calling getName() with any falsy value would not set hasName to true with your correct answer.

It would be helpful/essential for verification tools to have access to the Yul ASTs that are generated throughout compilation via Yul.

Add a mapping between custom accelerators and "nicknames" so that you can

Trainer(accelerator=‘nickname’)