Skip to content
Pytorch Mnist on Beaker
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
beaker_pytorch
Dockerfile
LICENSE
README.md
run.sh
spec.yml
spec2.yml
test.yml

README.md

MNIST Beaker Example

This is an example of how to train a Pytorch MNIST model in Beaker.

The mnist data is current already available on Beaker under marc/mnist.

Vanilla usage:

  1. install pytorch and torchvision
  2. set the python path:
export PYTHONPATH=.
  1. download the four MNIST files from Yann Lecun's website and put them somewhere
  2. run the main file
EPOCH=10; python beaker_pytorch/main.py \
    --data-dir /path/to/data \
    --output-dir /tmp/output \

Docker usage:

  1. build the docker image
docker build -t mnist .

Beaker usage

  1. Install Beaker
  2. You can choose to download the MNIST data above and create a new dataset or use the existing MNIST dataset named marc/mnist located here
  3. You can either rebuild and push the docker image above or use the existing Blueprint named marc/mnist

If you chose to rebuild use

beaker blueprint create -n mnist mnist

to create a blueprint. Please note that you may only have one Blueprint called mnist so may have to change this name.

  1. Run the experiment:
beaker experiment run -f spec.yml
You can’t perform that action at this time.