From 76c0153b6531e0be2c7c2d83fdd58415c7a6580f Mon Sep 17 00:00:00 2001 From: Jiaxiao Zheng Date: Wed, 19 Feb 2020 16:53:02 -0800 Subject: [PATCH] [Sample] update README (#3127) * [Sample] typo fix * update README --- samples/core/parameterized_tfx_oss/README.md | 14 +------------- 1 file changed, 1 insertion(+), 13 deletions(-) diff --git a/samples/core/parameterized_tfx_oss/README.md b/samples/core/parameterized_tfx_oss/README.md index 78d96f41e0c..c094d418e73 100644 --- a/samples/core/parameterized_tfx_oss/README.md +++ b/samples/core/parameterized_tfx_oss/README.md @@ -16,7 +16,7 @@ Please refer to inline comments for the purpose of each step in both samples. # Compilation * `parameterized_tfx_oss.py`: In order to successfully compile the Python sample, you'll need to have a TFX installation at -version 0.15.0 by running `python3 -m pip install tfx==0.15.0`. After that, under the sample dir run +version 0.21.0 by running `python3 -m pip install tfx==0.21.0`. After that, under the sample dir run `python3 parameterized_tfx_oss.py` to compile the TFX pipeline into KFP pipeline package. The compilation is done by invoking `kfp_runner.run(pipeline)` in the script. @@ -45,18 +45,6 @@ You can submit the compiled package to a KFP deployment and run it from the UI. The last step of the notebook the execution of the pipeline is invoked via KFP SDK client. Also you have the option to submit and run from UI manually. -## Caveats in `parameterized_tfx_oss.py` -This sample uses pipeline parameters in a TFX pipeline, which was not fully supported in TFX 0.15.0. -See [here](https://github.com/tensorflow/tfx/issues/362) for more details. In this sample, however, -the path to module file and path to data are parameterized. This is achieved by specifying those -objects `dsl.PipelineParam` and appending them to the `KubeflowDagRunner._params`. Then, -KubeflowDagRunner can correctly identify those pipeline parameters and interpret them as Argo -placeholder correctly when compilation. However, this parameterization approach is a hack and -we do not have plan for long-term support. Instead we're working with TFX team to support -pipeline parameterization using their -[RuntimeParameter](https://github.com/tensorflow/tfx/blob/592e05ea544d05f28d108ab74ebca70540854917/tfx/orchestration/data_types.py#L158). -You can check out the usage of `RuntimeParameter` in the notebook sample. - ### Known issues * This approach only works for string-typed quantities. For example, you cannot parameterize `num_steps` of `Trainer` in this way.