* Added component definitions to our components
Added the Kubeflow training sample pipeline that uses components
* Added the definition for "TFX - Data Validation"
* Added the definition for the "TFX - Analyze model" component
* Fixed bug in "ROC curve"
* Updated "Predict using TF on Dataflow"
* Updated "TFX - Data Validation"
* Updated the component definitions.
* Updated the pipeline to make the lines shorter and explicitly name the function parameters
* Changed the GCSPath type casing
* Added the definition for the "Kubeflow - Serve TF model" component
* Added the definition for the "Kubeflow - Launch StudyJob" component
* Removed all properties from GCPPath
This will confuse our users and make type checking worse, but Hongye and Ajay requested that.
`s/type: (\{GCSPath:.*?}})(.*)/type: GCPPath$2 # type: $1/g`
* Removed the usage of the ComponentStore
Now the samples are invalid until they're merged to master, but Hongye asked for that.
* Improve TFX Taxi Sample and Components.
- Confusion matrix and ROC components support flexible target.
- Use single mode instead of 4 different modes for TFDV, TFMA, TFT, Predicton.
- TFDV output's schema path directly.
- Add Confusion Matrix and ROC components to tf taxi sample.
* Follow up code review comments.
* migrate from the old repo
* fix bug: accidentally override tfma test
* add tfma test back
* add tfma back
* typo fix
* fix small typo
* if job fails, exit after logs are output
Ning
IronPan
Jiaxin Shan
hongye-sun
Alexey Volkov
Ning
Ajay Gopinathan
Jin Chi He
qimingj
Yang Pan
Pascal Vicaire