MLgym, a python framework for distributed machine learning model training in research.
Project description
MLgym
a python framework for distributed machine learning model training in research.
At its core, MLgym offers functionality to run gridsearches of Pytorch models at scale split over multiple GPUs and centrally store the results using DashifyML.
Futhermore, MLgym provides the following key features:
-
Reproducibility of results due to full experiment specification including dataset, preprocessing routines, model architecture, loss function, metrics and more within a single YAML config.
-
Component registry to register custom components with dependencies. For instance one can define a new preprocessing routine component. This component may depend on an iterator component, as specified in the experiment config. During runtime these components are instantiated on the fly.
-
Resume training after crash
-
Custom training routines, e.g., training with partially frozen network weights
-
Large scale, multi GPU training supporting Grid Search, Nested Cross Validation, Cross Validation
-
Reduced logging to reduce storage footprint of model and optimizer states
Please note, that at the moment this code should be treated as experimental and is not production ready.
Install
there are two options to install MLgym, the easiest way is to install it from the pip repository:
pip install mlgym
For the latest version, one can directly install it from source by cd into the root folder and then running
pip install src/
Usage
NOTE: This framework is still under heavy development and mainly used in research projects. It's most likely not free of bugs and interfaces can still change.
For usage see this example.
Copyright
Copyright (c) 2020 Max Lübbering, Rajkumar Ramamurthy
For license see: https://github.com/le1nux/mlgym/blob/master/LICENSE
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
File details
Details for the file mlgym-0.0.51.tar.gz.
File metadata
- Download URL: mlgym-0.0.51.tar.gz
- Upload date:
- Size: 43.1 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/3.7.1 importlib_metadata/4.10.0 pkginfo/1.8.2 requests/2.26.0 requests-toolbelt/0.9.1 tqdm/4.62.3 CPython/3.9.9
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 | 410ca602333d96309a0181ca04e5a7b6a415187e9bf7713c730f7c4e1b70a78b |
|
| MD5 | 35a1fbeac4121197df7fa9feba5596e5 |
|
| BLAKE2b-256 | 57f5a0e1c88aa4d2ca3e9fe15519989c6a269b1488b072e8b21e1bd6352d66ff |
File details
Details for the file mlgym-0.0.51-py3-none-any.whl.
File metadata
- Download URL: mlgym-0.0.51-py3-none-any.whl
- Upload date:
- Size: 62.6 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/3.7.1 importlib_metadata/4.10.0 pkginfo/1.8.2 requests/2.26.0 requests-toolbelt/0.9.1 tqdm/4.62.3 CPython/3.9.9
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 | 120a0d183f8f5b31dd202dbda57850695da57a1bb4c3143a7ec67b761901713e |
|
| MD5 | ae08f3277e70ce5fbaa067958d491e4a |
|
| BLAKE2b-256 | d872c35e51b1334509d6bbd41a9f42b053f10bf9b365b5ab8f4b7291de02d8a0 |
