Allow `ray up` to take URL as argument.

To get started with the autoscaler, we tell people to run a line like ray up ray/python/ray/autoscaler/aws/example-full.yaml. In practice, the steps are

1. Figure out that this means you have to clone the Ray repository.
2. Clone the Ray repository.
3. Figure out the new correct relative file path based on where you cloned it.
4. Run ray up ....

Instead, we could let people just copy and paste the following command:

ray up https://github.com/ray-project/ray/blob/master/python/ray/autoscaler/aws/example-full.yaml

In this blog post https://towardsdatascience.com/10x-faster-parallel-python-without-python-multiprocessing-e5017c93cce1, I posted benchmarks at the end that they could run with the autoscaler, but they still have to wget the relevant file. It’d be much easier if they could just use the URL.

Could also just write stuff like the following, but one line would be better.

wget https://raw.githubusercontent.com/ray-project/ray/master/python/ray/autoscaler/aws/example-full.yaml
ray up example-full.yaml

[robertnishihara]

@richardliaw the thing I like about the URL approach is that it is extensible, so if I write a blog post, I can host my own cluster yaml somewhere and then give people one command to run to reproduce the experiments.

[JosephTLucas]

Thanks! Yeah, I think I can do that. I’ll take a stab over the weekend and get you a PR.