What's CAC
This repository contains a library of Content-Aware Computing (CAC) by Fujitsu.
CAC is a software technology that aims at easy, high-speed, lightweight, and accurate deep learning processing.
Contents
1. Gradient-Skip
Gradient-Skip is an approach for CNNs to skip backward calculations for layers that enouch converged.
This reduces calculations in backward and communications of gradient.
You can use Gradient-Skip by simply replacing the optimizer with our SGD.
2. Automatic Pruner
Automatic Pruner is a pruning tool for neural networks, which can determine the pruning rate of each layer automatically.
3. Synchronous-Relaxation
Relaxed Synchronization technique removes slow processes from the group of distributed training and prevent limiting overall training speed due to slow processes.
Requirements
Python 3.6 or later
CUDA 10 or later
PyTorch 1.6 or later
Apex
Quick Start
Linux
git clone https://github.com/FujitsuLaboratories/CAC.git
cd CAC
python setup.py install
CAC will be registered in PyPI at a later date and will be able to pip install
pip install -v --disable-pip-version-check --no-cache-dir ./
1 maintainer
Release files for cac 1.0.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| cac-1.0.0.tar.gz | 21.1 kB | Details |
Release files / cac-1.0.0.tar.gz
| Download URL | cac-1.0.0.tar.gz |
|---|---|
| Size | 21.1 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
4f2d47453652d8346f0922d4ebaeed53b9aa06d70efc698a7dc9145f64468202
|
|
BLAKE2b-256 checksum How to use checksums |
6b183cb5760449c7f296f67ecb45fda21fbe341967d7cc58b4d488438bd77f15
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/3.4.2 importlib_metadata/4.8.1 pkginfo/1.5.0.1 requests/2.22.0 requests-toolbelt/0.9.1 tqdm/4.42.1 CPython/3.7.6
|
