https://github.com/typhoonzero/k8s_tensorflow_benchmark
my scripts to run tensorflow benchmark on Kubernetes GPU cluster
https://github.com/typhoonzero/k8s_tensorflow_benchmark
Last synced: 11 months ago
JSON representation
my scripts to run tensorflow benchmark on Kubernetes GPU cluster
- Host: GitHub
- URL: https://github.com/typhoonzero/k8s_tensorflow_benchmark
- Owner: typhoonzero
- Created: 2018-05-25T07:48:21.000Z (about 8 years ago)
- Default Branch: master
- Last Pushed: 2018-05-25T09:07:40.000Z (about 8 years ago)
- Last Synced: 2025-03-27T09:49:22.017Z (about 1 year ago)
- Language: Shell
- Size: 3.91 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# K8S Tensorflow Benchmark
This repo contains some example scripts to create a Kubernetes distributed
tensorflow benchmark job and get the results.
NOTE: these yamls are generated using a script from https://github.com/PaddlePaddle/Paddle/blob/develop/benchmark/fluid/kube_gen_job.py
, you can update this yaml files if needed.
* To build the docker image, you have to download file `libcudnn.so.7.0.5` first and put it under cloned directory.
See [here](https://developer.nvidia.com/cudnn)
to get the download link, then you can run the following command to build the docker image:
```bash
docker built -t [image:tag] .
```
* To start the job on Kubernetes, run:
```bash
kubectl create -f tf_local.yaml
# or
kubectl create -f tf_dist.yaml
```