|
1 | | -# ElasticDL Client: Submit ElasticDL Job to Kubernetes |
| 1 | +# ElasticDL Client: Submit ElasticDL Job to Kubernetes |
2 | 2 |
|
3 | 3 | ## Prepare Model Definition |
4 | 4 |
|
@@ -29,13 +29,13 @@ For demonstration purposes, we use the data stored on `elasticdl:ci` Docker imag |
29 | 29 | First we build all development Docker images, which include `elasticdl:ci` image: |
30 | 30 |
|
31 | 31 | ```bash |
32 | | -elasticdl/docker/build_all.sh |
| 32 | +export TRAVIS_BUILD_DIR=$PWD |
| 33 | +bash scripts/travis/build_images.sh |
33 | 34 | ``` |
34 | 35 |
|
35 | | -Add `-gpu` if you want to build images with GPU support: |
36 | | -```bash |
37 | | -elasticdl/docker/build_all.sh -gpu |
38 | | -``` |
| 36 | +By default, the above script builds images with TensorFlow CPU image as the base |
| 37 | +image. If you want to switch to other images, for example, Python, Ubuntu, or |
| 38 | +TensorFlow GPU image, please edit `elasticdl/docker/Dockerfile`. |
39 | 39 |
|
40 | 40 | Submit training job (make sure you have packages `kubernetes` and `docker` installed in your running environment): |
41 | 41 |
|
@@ -101,7 +101,7 @@ python -m elasticdl.python.elasticdl.client train \ |
101 | 101 |
|
102 | 102 | The difference is that we add a new argument `cluster_spec` which points to a cluster specification file. |
103 | 103 | The cluster specification module includes a `cluster` component, and ElasticDL will invoke function |
104 | | -`cluster.with_cluster(pod)` to add extra specifications to the |
| 104 | +`cluster.with_cluster(pod)` to add extra specifications to the |
105 | 105 | [pod](https://github.com/kubernetes-client/python/blob/master/kubernetes/docs/V1Pod.md) and invoke function |
106 | 106 | `cluster.with_service(service)` to add extra specifications to the [service](https://github.com/kubernetes-client/python/blob/master/kubernetes/docs/V1Service.md). |
107 | 107 | Here is an example that assigns labels `"app": "elasticdl"` to the `pod` and `service`. Users can implement more customized configurations |
|
0 commit comments