https://github.com/feifeibear/89757
https://github.com/feifeibear/89757
Last synced: 5 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/feifeibear/89757
- Owner: feifeibear
- Created: 2018-05-28T22:25:24.000Z (over 7 years ago)
- Default Branch: master
- Last Pushed: 2018-07-21T03:06:38.000Z (over 7 years ago)
- Last Synced: 2025-07-04T17:50:13.091Z (6 months ago)
- Language: Python
- Homepage:
- Size: 4.51 MB
- Stars: 2
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
This code support training CNN with GC for cifar10 and Imagenet dataset
##How to run this code
### Train Cifar10
if you computer do not connect to the internet
1. download cifar10 to ./dataset/pytorch
mkdir -p ./dataset/pytorch/cifar10
download the follows data to that directory
https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz
2. Modify the submit scripts
run on GPU cluster, where each node has one card
export USE_CLUSTER=use_cluster
set the number of node to run
export NUM_NODE=2
use bsub instead of mpirun
3. run code
To test GC, you should run
./submit_dist_dgc.sh
To test SGD, you should run
./submit_dist_sgd.sh