DASH-pytorch

DASH (wip)

Implementation of DASH, Warm-Starting Neural Network Training in Stationary Settings without Loss of Plasticity

Install

$ pip install DASH-pytorch

Usage

import torch
from torch.nn import Linear

from DASH.DASH import AdamW

net = Linear(10, 5)
optim = AdamW(net.parameters(), lr = 3e-4)

loss = net(torch.randn(10)).sum()
loss.backward()

optim.step()
optim.zero_grad()

optim.shrink_params()
optim.clear_grad_ema()

Citations

@misc{shin2024dashwarmstartingneuralnetwork,
    title   = {DASH: Warm-Starting Neural Network Training in Stationary Settings without Loss of Plasticity}, 
    author  = {Baekrok Shin and Junsoo Oh and Hanseul Cho and Chulhee Yun},
    year    = {2024},
    eprint  = {2410.23495},
    archivePrefix = {arXiv},
    primaryClass = {cs.LG},
    url     = {https://arxiv.org/abs/2410.23495}, 
}

Metadata

Release files for DASH-pytorch 0.0.3

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for DASH-pytorch 0.0.3
File	Size	Uploaded
dash_pytorch-0.0.3.tar.gz	105.9 kB	Oct 17, 2025	Details

Built distribution (wheel)

Table of built distributions (wheels) for DASH-pytorch 0.0.3
File	Interpreter	ABI	Platform	Reset
dash_pytorch-0.0.3-py3-none-any.whl 5.3 kB Oct 17, 2025	Python 3	none	any	Details

Total release size: 111.2 kB

Release files / dash_pytorch-0.0.3.tar.gz

Download URL	dash_pytorch-0.0.3.tar.gz
Size	105.9 kB
Tags	Source
SHA-256 checksum How to use checksums	`3f3c2d3d292ed8133dcdc348625a826571fd3533968ce022800ef2d1167a456c`
BLAKE2b-256 checksum How to use checksums	`44ed137857f95d0af85f1e7436818803310c6bde0fa98eb120506e228beb2ef9`
Upload date	Oct 17, 2025
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`twine/6.2.0 CPython/3.9.24`

Release files / dash_pytorch-0.0.3-py3-none-any.whl

Download URL	dash_pytorch-0.0.3-py3-none-any.whl
Size	5.3 kB
Tags	Python 3
SHA-256 checksum How to use checksums	`65e03be63eae1b5a72a688c3538e9d1780e473ed3505801b8dee257e6cb5e124`
BLAKE2b-256 checksum How to use checksums	`12f5c02ed30a5d87ea9efb477749d8be6fa565828afa90d093831be48671dda7`
Upload date	Oct 17, 2025
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`twine/6.2.0 CPython/3.9.24`