https://github.com/gaocegege/kubeflow-benchmarks
Benchmarks for Kubeflow
https://github.com/gaocegege/kubeflow-benchmarks
Last synced: 2 months ago
JSON representation
Benchmarks for Kubeflow
- Host: GitHub
- URL: https://github.com/gaocegege/kubeflow-benchmarks
- Owner: gaocegege
- Created: 2018-09-03T06:04:40.000Z (over 6 years ago)
- Default Branch: master
- Last Pushed: 2020-10-07T09:07:02.000Z (over 4 years ago)
- Last Synced: 2025-02-08T15:48:04.792Z (4 months ago)
- Language: HTML
- Size: 266 KB
- Stars: 2
- Watchers: 2
- Forks: 1
- Open Issues: 2
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# kubeflow-benchmarks
## TensorFlow on Kubernetes
### Local
- Native and dockerized cnn benchmark (1 GPU)
- Native and dockerized cnn benchmark (2 GPUs)
```bash
$ docker build -t kubeflow/benchmarks .
# Dockerized
$ docker run -v $(pwd):/kubeflow-benchmarks kubeflow/benchmarks ./scripts/single.sh
$ docker run -v $(pwd):/kubeflow-benchmarks kubeflow/benchmarks ./scripts/double.sh
# Native
$ ./scripts/single.sh
$ ./scripts/double.sh
```
### Distributed
#### PS/Worker distribution strategy
- Native and dockerized in one machine
- Native and dockerized in different machines
- Kubernetes default-scheduler vs kube-arbitrator vs ideal-scheduler
#### Mirrored distribution strategy
- Native and dockerized in one machine
#### Collective AllReduce strategy
- Native and dockerized in one machine
- Native and dockerized in different machines
- Kubernetes default-scheduler vs kube-arbitrator vs ideal-scheduler
#### Uber Horovod (Optional)
- Native and dockerized in one machine
- Native and dockerized in different machines
- Kubernetes default-scheduler vs kube-arbitrator vs ideal-scheduler
## PyTorch on Kubernetes
TBA