Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/pnkraemer/experiments-lanczos-adjoints
https://github.com/pnkraemer/experiments-lanczos-adjoints
Last synced: 15 days ago
JSON representation
- Host: GitHub
- URL: https://github.com/pnkraemer/experiments-lanczos-adjoints
- Owner: pnkraemer
- License: mit
- Created: 2023-12-19T07:39:01.000Z (about 1 year ago)
- Default Branch: main
- Last Pushed: 2024-05-28T07:36:14.000Z (7 months ago)
- Last Synced: 2024-05-28T21:35:55.324Z (7 months ago)
- Language: Python
- Size: 68.4 MB
- Stars: 0
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# experiments-lanczos-adjoints
This repository contains the experiment code for the paper
> **Gradients of functions of large matrices.**
> _Nicholas Krämer, Pablo Moreno-Muñoz, Hrittik Roy, Søren Hauberg._
> 2024
> Preprint on arXiv:2405.17277.
[Here is a link to the preprint](https://arxiv.org/abs/2405.17277).
Please cite the paper as follows:
```bibtex
@article{kraemer2024gradients,
title={Gradients of functions of large matrices},
author={Kr\"amer, Nicholas and Moreno-Mu\~noz, Pablo and Roy, Hrittik and Hauberg S\o{}ren},
journal={arXiv preprint arXiv:2405.17277},
year={2024}
}
```
_**A more accessible version of the code provided in here can be found [in the matfree library](https://github.com/pnkraemer/matfree). Install it via ``pip install matfree``.
Unless you are looking for specific experiment configurations, you likely want use matfree, not this repository here.**_
## Installation
Before installing this project,
and [after creating & activating your virtual environment](https://realpython.com/python-virtual-environments-a-primer/),
you must install JAX yourself because CPU and GPU backends require different installation commands.
See [here](https://jax.readthedocs.io/en/latest/installation.html) for instructions.
For the small examples, `pip install jax[cpu]` will suffice.
For the bigger experiments, a GPU is helpful.
Then, install the project via
```commandline
pip install .
```
To install the development-related dependencies, run
```commandline
pip install ".[dev]"
```
To install the experiment-related dependencies, first make sure
that JAX and pytorch are installed (JAX has been mentioned above;
for PyTorch, see [here](https://pytorch.org/), then run
```commandline
pip install ".[experiments]"
```
To install everything and in editable mode (recommended!),
first install JAX and PyTorch, and then run
```commandline
pip install -e ".[dev,experiments]"
```
## Running dev-related code
Run tests via
```
pytest
```
or use the pre-defined script
```commandline
make test
```
## Pre-commit hook
after activating your virtual environment, run
```commandline
pre-commit install
```
you only do this once! not every time you activate the environment.
The pre-commit hook is useful, because it implies that we do not accidentally (auto)format each others' code, which would lead to nasty merge conflicts.
## Running code and saving results/figures
Ideally, place experiment scripts into a dedicated subdirectory of experiments/...
E.g., GP related code into experiments/applications/gaussian_process.
Save figures and results in (maybe matching) directories in results/ or figures/.
This could be easier with `matfree_extensions.exp_util.create_matching_directory()`; see the existing scripts for how to use it.
(requires running all scripts from the root, i.e. from where pyproject.toml is.)
## Using code
Look at tests for how to use the library functions.