https://github.com/nhynes/em
Tool for managing deep learning experiments
https://github.com/nhynes/em
deep-learning git pytorch
Last synced: over 1 year ago
JSON representation
Tool for managing deep learning experiments
- Host: GitHub
- URL: https://github.com/nhynes/em
- Owner: nhynes
- License: mit
- Created: 2017-03-30T16:03:28.000Z (about 9 years ago)
- Default Branch: master
- Last Pushed: 2017-12-22T21:48:27.000Z (over 8 years ago)
- Last Synced: 2025-03-18T20:06:40.858Z (over 1 year ago)
- Topics: deep-learning, git, pytorch
- Language: Python
- Size: 25.4 KB
- Stars: 13
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# experiment manager
`em` is a tool that facilitates deep learning experimentation using a [Git worktree](https://git-scm.com/docs/git-worktree) workflow.
The use case is when you want to test a number of small changes that are not worth implementing as into full-fledged options but must still be recorded, nonetheless.
`em` allows this by creating a separate worktree for each experiment; this work tree is essentially a snapshot of the configuration at the time the experiment was run.
Additionally, since each experiment has its own branch, interesting options can be merged back into the main branch.
For instance,
```
em proj test_proj
cd test_proj
git add -A && git commit -m "Initial commit."
echo "print('hello, world!')" > main.py
em run testing # hello world!
em show testing # {'created': 1490891521.6767356, 'status': 'completed'}
em clean testing
```
## Requirements and Installation
You will need
* Python >= 3.6
* libgit2 >= 0.26
* pygit2 >= 0.26
* `pip install daemon`
## Usage
`em --help`