Open vict0rsch opened 7 months ago
Would be great to specify an identifier (like a project name or something) to map a project to a path, possibly per cluster. Something like
# ~/.milatools.yaml myproject: mila: /home/mila/s/schmidtv/project/code/repos/repoName beluga: /home/vsch/scratch/project/repos/repoName myproject2: mila: /home/mila/s/schmidtv/thatthing/foo/proj2 beluga: /home/vsch/scratch/thatotherthing/bar/proj2-github
and then specify an identifier or project with -p myproject2 (or -i myproject2)
-p myproject2
-i myproject2
You could even push this feature further to be more complex and precise with alloc defaults
# ~/.milatools.yaml myproject: mila: path: /home/mila/s/schmidtv/project/code/repos/repoName alloc: cpus: 2 memory: 12G beluga: path: /home/vsch/scratch/project/repos/repoName myproject2: mila: path: /home/mila/s/schmidtv/thatthing/foo/proj2 alloc: memory: 16G beluga: path: /home/vsch/scratch/thatotherthing/bar/proj2-github
And potentially overwrite defaults with something like --remember-alloc
--remember-alloc
And milatools would always create a config file with an example structure, either in comments or as an unused dummy project, if it does not already exist
Would be great to specify an identifier (like a project name or something) to map a project to a path, possibly per cluster. Something like
and then specify an identifier or project with
-p myproject2
(or-i myproject2
)You could even push this feature further to be more complex and precise with alloc defaults
And potentially overwrite defaults with something like
--remember-alloc