abhinavsp0730/kedro-wdbc-tf

Example for using Kedro for Tensorflow Models.

In this example I've used wdbc(Breast Cancer Dataset) and used Tensorflow to train the model against the wdbc dataset. All the Machine Learning WorkFlow is created by kedro

This Project is Created in PyCon India 2020 Devsprint.

Machine Learning WorkFlow visualization created by kedro-viz.

Important Mention

This project wouldn't be possible without the constanst support of Lais Carvalho

kedro-wdbc-tf's Issues

Error on `kedro run`

Hello Abhinav :)

I came from the Zulip channel straight into your repo 📦

Now, the reason why you are getting the kedro.framework.context.context.KedroContextError: Pipeline contains no nodes is because, well... your project contain no pipelines. It seems like you do have them, actually but not on the folder where the runner will look for it. Hence, when the runner.py passes by to record the dependencies and relations between your nodes (pipeline building) it cannot find them and returns that error.

I can see you have two sets of files and folders so I think you might have generated a kedro project with a version earlier than 0.16.5 and then updated. Which is no problem at all, we just need to change a small thing.

Since Kedro 0.16.5 uses the hooks.py file to read in pipeline entries, we need to transfer you data pipelines to it.

To do so, do the following:

1. On /src//hooks.py, under the class ProjectHooks add your pipelines:

        return {"__default__": Pipeline([
              node(
                func=create_features,
                inputs="wdbc",
                outputs="features",
                name="creating_features",
            ),
            node(
                func=create_labels,
                inputs="wdbc",
                outputs="labels",
                name="creating_labels",
            ),
])}

Save the file and type kedro run again. Let me know if this solves your issue :)

In any case, we're more than happy to debug this bad boy at Kedro Community as well :)

Be good and keep me posted.

Thanks for contributing!