Code cleaning
Regarding the general cleaning task of Issue #1, I open this issue to list every small inconsistency in the code we need to fix:
-
draw_param_set
should be abstracted -> solved with MR !27 (merged). -
the learning
parameter should be removed from all config files (e.g. here). -
increase launcher verbosity for easier CI evaluation -
All Python/CMake headers should be replaced in favor of the one in the launcher files. -
the .gitignore
file should be edited. -
the .gitlab-ci
file should be purged of unnecessary commented parts. -
commented core of functions should be removed from the server files. Reorganize inheritance to remove duplicate code (i.e. make a TorchServer
). -
outdated and commented parts of the main CMake
file should be removed. -
nserver.py
should be renamedserver.py
or something like that. -
plot scripts should be added to the heat-pde-sa, heat-pde-dl and lorenz use cases (see existing files here and there). Doc update was done with MR !42. -
clean client.sh
andmodules.melissa.sh
files (cf. doc description, make one modules file for both server/client). -
there should be a default output-directory (cf. doc description). -
remove outdated and unused logger.py
(seemelissa/server/logger.py
). -
Unnecessary return
(e.g. here) should be removed. -
Duplicate assert False
andraise
should be consistent (e.g. here). I think theraise
is to be preferred because it can be catch if necessary. Similarly thesys.exit
should be make consistent.raise
would be once again the way to go. -
Simplify function names/order launcher run()
and_handle_filed()
-
Move server main to within server folder. -
Add setup.py for pip install of melissa? -
Add flexibility of relative/absolute paths for project_dir and config_name. -
Remove redundant DL params from config/example servers and make examples online. -
Convert repo name from "Melissa Combined" to "Melissa." Rename old melissa repo to "Melissa Classic" (or something similar). Change all mentions of Melissa Combined to Melissa throughout source code (there are a lot, docs, CI, etc.). -
simplify .gitlab-ci.yml, remove all pip installs, and unnecessary path exports (update docker file to automatically handle paths/installs). -
Put all client.X.sh scripts, *.err, and *.out scripts in their own folders to make output_dir manageable for large jobs -
Make parameters case consistent