library for deep learning and privacy preserving deep learning
Project description
hideandseek
deep learning and privacy preserving deep learning library.
Why use hideandseek
?
- Run multiple deep learning experiements in parallel on multiples GPUs
- Design and analyze experiments scientifically by modifying variables (hydra)
- Modularized machine learning pipeline allows using the same script for all types of experiments
- The same training code can be run in privacy preserving setting by minimal modifications
Currently integrating from experiment codes. (30.10.2021.)
import torch
from omegaconf import OmegaConf
import hideandseek as hs
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
cfg = OmegaConf.load('config.yaml') # omegaconf.OmegaConf.DictConfig object
model = DNN() # torch.nn.Module object
train_dataset = dataset # torch.utils.data.Dataset object
kwargs = {
'model': model,
'dataset': train_dataset,
'cfg_train': cfg,
'criterion': criterion,
}
node = hs.Node(**kwargs)
model.to(device)
node.step(local_T=20, horizon='epoch') # trains for 20 epochs
# node.step(local_T=1000, horizon='step') # trains for 1000 steps
model.cpu()
node.save()
test_results = hs.eval.test(node)
scores = hs.eval.scores(test_results)
To do
- Migrate modules from experiment codes
- GUI for generating experiment scripts when conducting variable sweeps
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
hideandseek-0.1.1.3.tar.gz
(10.9 kB
view hashes)
Built Distribution
Close
Hashes for hideandseek-0.1.1.3-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 779b586db112463561b126b613fa6220cd869619474a5c67da5b1e4c3e00a67e |
|
MD5 | d87fb650bbe54f17caf757b45aa1b720 |
|
BLAKE2b-256 | 6dc774ece113dc3b5be2477b15a4dee6a4d4c4294f1d7327ebec46e655f1228f |