Set an env variable to indicate that it's a FloydHub machine [Solved]

[kul]

Our code could use this to identify that it's a FloydHub machine it is running on (mainly for jupyter notebooks)

This would be helpful to set the input/output paths so that the code is portable between FH and our local machines.

[scottdbrooks]

Workaround:

floyd run "export ENVVAR=/your/paths; python your-code.py"

[superzadeh]

It would be good to have one defined automatically in all environments (tensorflow, pytorch etc.)

[redeipirati]

Hi @superzadeh,

Here's another possible workaround, you can use the value of the cwd (current working directory). Every time you run a Job or work inside a Workspace, the default cwd is /floyd/home. Here's a template code that you can use to discriminate if you are local or inside a FloydHub environment:

import os

IS_FLOYD_ENV = True if os.getcwd() == '/floyd/home' else False

Hope that helps.

[superzadeh]

Thanks @redeipirati, that's super useful! We'll roll with this for now, however it would be nice to have this IS_FLOYD_ENV set automatically.

Travis does a very good job at providing a sensible amount of environment variables: https://docs.travis-ci.com/user/environment-variables#default-environment-variables

The way we use FloydHub with jobs is actually very close to a CI: we change a script, then run the related job on FloydHub, which spins up a container configured with specific frameworks (similar to specifying the language in Travis) and executes our task within it.

[houqp]

Hi @superzadeh,

Sorry for the bad documentation. We actually already have a env var set for this purpose. For all floydhub jobs and workspaces, we set FLOYDHUB=1. We will work on updating our doc to reflect this.

Thanks,
QP