Shiweiliuiiiiiii/In-Time-Over-Parameterization
[ICML 2021] "Do We Actually Need Dense Over-Parameterization? In-Time Over-Parameterization in Sparse Training" by Shiwei Liu, Lu Yin, Decebal Constantin Mocanu, Mykola Pechenizkiy
This project helps machine learning practitioners develop neural networks that perform well without requiring extensive computational resources during training. It takes in standard deep learning models and datasets and outputs highly accurate, 'sparse' models that are more efficient to train. Data scientists and ML engineers working on large-scale image classification or similar tasks would find this particularly useful.
No commits in the last 6 months.
Use this if you need to train deep neural networks that achieve state-of-the-art accuracy but are constrained by computational resources or want to reduce training time and energy consumption.
Not ideal if you are working with small datasets or simpler models where the benefits of sparse training are less critical, or if you prefer to stick with traditional 'dense' network training for maximum expressibility without considering efficiency.
Stars
45
Forks
7
Language
Python
License
—
Category
Last pushed
Nov 11, 2023
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/Shiweiliuiiiiiii/In-Time-Over-Parameterization"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
InterDigitalInc/CompressAI
A PyTorch library and evaluation platform for end-to-end compression research
quic/aimet
AIMET is a library that provides advanced quantization and compression techniques for trained...
tensorflow/compression
Data compression in TensorFlow
baler-collaboration/baler
Repository of Baler, a machine learning based data compression tool
thulab/DeepHash
An Open-Source Package for Deep Learning to Hash (DeepHash)