kubeflow
/
examples
mirror of https://github.com/kubeflow/examples.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
examples/pytorch_mnist/training/ddp/mnist
History
David Sabater Dinter 152c38b386 [mnist_pytorch] Optimise build and switch backend from MPI to GLOO (#480) 
* Refactor Python module:
- Replace MPI by GLOO as backend to avoid having to recompily Pytorch
- Replace DistributedDataParallel() class with official version when using GPUs
- Remove unnecessary method to disable logs in workers
- Refactor run()

* Simplify Dockerfile by using Pytorch 0.4 official image with Cuda and remove mpirun call
 		2019-01-16 11:38:52 -08:00
..
cpu revert back removed v1alpha2 yaml manifests (#475) 2019-01-14 17:08:29 -08:00
gpu revert back removed v1alpha2 yaml manifests (#475) 2019-01-14 17:08:29 -08:00
Dockerfile.traincpu [mnist_pytorch] Optimise build and switch backend from MPI to GLOO (#480) 2019-01-16 11:38:52 -08:00
Dockerfile.traingpu [mnist_pytorch] Optimise build and switch backend from MPI to GLOO (#480) 2019-01-16 11:38:52 -08:00
build_image.sh E2E Pytorch mnist example (#274) 2018-11-18 14:24:43 -08:00
mnist_DDP.py [mnist_pytorch] Optimise build and switch backend from MPI to GLOO (#480) 2019-01-16 11:38:52 -08:00