https://github.com/lucidrains/grokfast-pytorch
Explorations into the proposal from the paper "Grokfast, Accelerated Grokking by Amplifying Slow Gradients"
https://github.com/lucidrains/grokfast-pytorch
artificial-intelligence deep-learning grokking
Last synced: about 1 year ago
JSON representation
Explorations into the proposal from the paper "Grokfast, Accelerated Grokking by Amplifying Slow Gradients"
- Host: GitHub
- URL: https://github.com/lucidrains/grokfast-pytorch
- Owner: lucidrains
- License: mit
- Created: 2024-06-15T15:27:33.000Z (about 2 years ago)
- Default Branch: main
- Last Pushed: 2024-12-22T16:48:15.000Z (over 1 year ago)
- Last Synced: 2025-03-29T00:06:04.451Z (over 1 year ago)
- Topics: artificial-intelligence, deep-learning, grokking
- Language: Python
- Homepage:
- Size: 166 KB
- Stars: 98
- Watchers: 3
- Forks: 7
- Open Issues: 4
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
## Grokfast - Pytorch (wip)
Explorations into "Grokfast, Accelerated Grokking by Amplifying Slow Gradients", out of Seoul National University in Korea. In particular, will compare it with NAdam on modular addition as well as a few other tasks, since I am curious why those experiments are left out of the paper. If it holds up, will polish it up into a nice package for quick use.
The official repository can be found here
## Install
```bash
$ pip install grokfast-pytorch
```
## Usage
```python
import torch
from torch import nn
# toy model
model = nn.Linear(10, 1)
# import GrokFastAdamW and instantiate with parameters
from grokfast_pytorch import GrokFastAdamW
opt = GrokFastAdamW(
model.parameters(),
lr = 1e-4,
weight_decay = 1e-2
)
# forward and backwards
loss = model(torch.randn(10))
loss.backward()
# optimizer step
opt.step()
opt.zero_grad()
```
## Todo
- [ ] run all experiments on small transformer
- [ ] modular addition
- [ ] pathfinder-x
- [ ] run against nadam and some other optimizers
- [ ] see if `exp_avg` could be repurposed for amplifying slow grads
- [ ] add the foreach version only if above experiments turn out well
## Citations
```bibtex
@inproceedings{Lee2024GrokfastAG,
title = {Grokfast: Accelerated Grokking by Amplifying Slow Gradients},
author = {Jaerin Lee and Bong Gyun Kang and Kihoon Kim and Kyoung Mu Lee},
year = {2024},
url = {https://api.semanticscholar.org/CorpusID:270123846}
}
```
```bibtex
@misc{kumar2024maintaining,
title = {Maintaining Plasticity in Continual Learning via Regenerative Regularization},
author = {Saurabh Kumar and Henrik Marklund and Benjamin Van Roy},
year = {2024},
url = {https://openreview.net/forum?id=lyoOWX0e0O}
}
```
```bibtex
@inproceedings{anonymous2024the,
title = {The Complexity Dynamics of Grokking},
author = {Anonymous},
booktitle = {Submitted to The Thirteenth International Conference on Learning Representations},
year = {2024},
url = {https://openreview.net/forum?id=07N9jCfIE4},
note = {under review}
}
```