https://github.com/petuum/tuun
Hyperparameter tuning via uncertainty modeling
https://github.com/petuum/tuun
Last synced: 9 months ago
JSON representation
Hyperparameter tuning via uncertainty modeling
- Host: GitHub
- URL: https://github.com/petuum/tuun
- Owner: petuum
- License: apache-2.0
- Created: 2020-06-23T23:33:29.000Z (about 6 years ago)
- Default Branch: master
- Last Pushed: 2024-05-03T19:39:07.000Z (about 2 years ago)
- Last Synced: 2024-05-08T00:19:19.961Z (about 2 years ago)
- Language: Python
- Homepage:
- Size: 60.9 MB
- Stars: 44
- Watchers: 4
- Forks: 3
- Open Issues: 4
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
**Tuun** is a toolkit for efficient hyperparameter tuning via uncertainty
modeling, with a focus on flexible model choice, scalability, and use in
distributed settings.
## Installation
Tuun requires Python 3.6+. To install all dependencies for development, clone and `cd`
into this repo, and run:
```
$ pip install -r requirements/requirements_dev.txt
```
For the full functionality of Tuun, a [Stan](https://mc-stan.org/) model must also be
compiled (this takes roughly 1 minute) by running:
```
$ python tuun/probo/models/stan/compile_models.py -m gp_distmat_fixedsig
```
## Quick Start
**Note:** Tuun is still in the early stages and the following API and functionality may
undergo changes.
[Here](examples/quick_start/minimize_01.py) is a minimal working example, which uses
Tuun to optimize a function over a one-dimensional Euclidean search space with bounds
[-5, 5].
```python
from tuun.main import Tuun
# instantiate Tuun
tu = Tuun()
# set search space
search_space = ('real', [-5, 5])
tu.set_config_from_list(search_space)
# define function to optimize
f = lambda x: x[0] ** 4 - x[0] ** 2 + 0.1 * x[0]
# minimize function over search space
result = tu.minimize_function(f, 20)
```
This should find a minima at roughly: 𝑥\*=[−0.73], where 𝑓(𝑥\*)=−0.32. See [this docs
page](https://petuum.github.io/tuun/search_space.html) for more details on defining
different search spaces for Tuun.
Tuun also allows for fine-grained configuration of individual components and search
spaces, as shown in the [example](examples/quick_start/minimize_11.py) below.
```python
from tuun.main import Tuun
config = {
# configure tuning backend
'backend': 'probo',
# configure model
'model_config': {'name': 'simplegp', 'ls': 3.0, 'alpha': 1.5, 'sigma': 1e-5},
# configure acquisition function
'acqfunction_config': {'name': 'default', 'acq_str': 'ei'},
# configure acquisition optimizer
'acqoptimizer_config': {'name': 'default', 'max_iter': 200},
# configure domain
'domain_config': ('real', [-5, 5]),
}
tu = Tuun(config)
f = lambda x: x[0] ** 4 - x[0] ** 2 + 0.1 * x[0]
result = tu.minimize_function(f, 20)
```
This should also find a minima at roughly: 𝑥\*=[−0.73], where 𝑓(𝑥\*)=−0.32. See [this
docs page](https://petuum.github.io/tuun/configure.html) for more details on possible
configurations.
[See here](examples/quick_start/) for a larger set of quick start examples for Tuun.
## Use in NNI
Tuun can be used as a custom tuner in [NNI (neural network
intelligence)](https://github.com/microsoft/nni), which allows for visualization and
experiment management. See [this docs page](https://petuum.github.io/tuun/nni.html) for
more details, and [this directory](examples/nni_simple_2d) for a minimal example.
## Examples
See a few examples of Tuun [here](examples/), including the setup for various benchmark
functions, different types of search spaces, and use within NNI.
The plots below show a couple examples of Tuun, along with other tuning algorithms, on
benchmark functions.
## Affiliations
Tuun is part of the [CASL project](http://casl-project.ai/).
Companies and universities using and developing Tuun.
