A package for performing hyperparameter tuning with the SLURM scheduling system.
Project description
slune (slurm + tune!)
A super simplistic way to perform hyperparameter tuning on a cluster using SLURM. Will submit a seperate job script for each run in the tuning and store metrics in a csv and then allow you to easily query for the best hyperparamters based on metric.
Currently very much in early stages, first things still to do:
- Add ability to read results, currently can only submit jobs and log metrics during tuning.
- Refine class structure, ie. subclassing, making sure classes have essential methods, what are the essential methods and attributes? etc.
- Refine package structure and sort out github actions like test coverage, running tests etc.
- Add interfacing with SLURM to check for and re-submit failed jobs etc.
- Add more tests and documentation.
- Add some more subclasses for saving job results in different ways and for different tuning methods. Although the idea for this package is to keep it ultra bare-bones and make it easy for the user to mod and add things themselves to their liking.
To run tests use:
python -m unittest discover -s . -p 'test_*.py'
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
slune-lib-0.0.1.tar.gz
(7.0 kB
view hashes)
Built Distribution
Close
Hashes for slune_lib-0.0.1-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | ac5fdc2fd234af6201da3d3cb2baa5a331ace5bba093f5074959d4d83ec3d583 |
|
MD5 | e471d90cc0713634ed63661781ec5e46 |
|
BLAKE2b-256 | 2e366ddf43efaab397f79d0648c8fa65e6abb1542cca2b3734abc99568f4adb3 |