A package with utilities for managing and running machine learning projects
Project description
# ML Project Manager
The intent of this project is to provide a quick and easy to use framework to run machine learning experiments in a systematic way, while keeping track of all the important details that are necessary for reproducibility.
## Installation This project is uploaded to the Python Package Index, so you can simply run the following command: python3 -m pip install mlproj_manager
## Usage Here is a quick list of steps to create and run a new experiment:
1. Write a python script with a class that is a child of the Experiment abstract class in ./mlproj_manager/experiments/abstract_experiment.py. See ./examples/non_stationary_cifar_example for an example. 2. Register the experiment using the command python -m mlproj_manager.experiments.register_experiment with the arguments –experiment-name followed by a named of your choosing, –experiment-path followed by the path to the script created in step 1, and –experiment-class-name followed by the name of the class defined in the script created in step 1. 3. Create a config.json file for your experiment that contains all the relevant details for running the experiment. See ./examples/non_stationary_cifar_example/config_files/backprop.json for an example. 4. Finally, run the experiment using the command python -m mlproj_manager.main with the arguments –experiment-name followed by the experiment name used in step 2, –experiment-config-path followed by the path to the config file created in step 3, –use-slurm (optional) to indicate whether to schedule the experiment using slurm, and –slurm-config-path (required only if using slurm) followed by the path to a similar file as the one created for step 3 but with parameters relevant to the slurm scheduler.
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
Built Distribution
Hashes for mlproj_manager-0.0.7-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 9b3867e1d214bbf69697b40f4cc8fca3ee467eacf376894abef341103ed614c0 |
|
MD5 | 626a5fa9c87125cd0780a7fb69df0a2e |
|
BLAKE2b-256 | a8d0b29b1a2de47ed28c4068473cf007a0d60433d0c00605078afae8c6ec45bc |