How to plot on kedro mlflow ui x1=array/list/dict and y1=array/list/dict? https://stackoverflow.com/questions/71713641/how-to-plot-on-kedro-mlflow-ui-x1-array-list-dict-and-y1-array-list-dict

Regarding modular pipelines: if my pipeline persists intermediate datasets (i.e. they're listed as outputs of intermediate nodes rather than the last one), can I override those outputs as well, or is this only possible for the outputs of the final node?

Good question - as they're both inputs and outputs I'm not sure. You could try? I think you may need a to override them in both settings if it's possible

In truth the actual advice is to break your pipeline into smaller pieces or remove the persistence

Thanks, I'll try to refactor the pipeline then

You can definitely override those outputs. As per this docstring https://kedro.readthedocs.io/en/latest/kedro.pipeline.modular_pipeline.pipeline.html#kedro.pipeline.modular_pipeline.pipeline,

outputs

means "free outputs or intermediate outputs" and

inputs

means "just free inputs".

But the general point about not persisting intermediate datasets unless there's a good reason to do so is still very valid - people tend to persist more things than they really need to from what I've seen

Thanks, point taken. I do need said intermediate datasets for later on, but it's definitely something to keep in mind at all times

Hi In new release 0.18.0 when I'm trying to uncomment CONFIG_LOADER_CLASS=TemplatedConfigLoader in settings.py I get an error: "dynaconf.validator.ValidationError: Invalid value

kedro.config.templated_config.TemplatedConfigLoader

received for setting

CONFIG_LOADER_CLASS

. It must be a subclass of

kedro.config.config.ConfigLoader

. " Is it a bug?

Hi @Elzoschka , we are aware of this issue. There is aa adhoc fix in the link, we will address this issue very soon. https://github.com/kedro-org/kedro/issues/1402

Thanks!

According to the modular pipeline docs, parameters can be remapped to new names using a dict (https://kedro.readthedocs.io/en/stable/kedro.pipeline.modular_pipeline.pipeline.html), but I'm not sure I'm getting it - the dict I'm passing to the

pipeline

wrapper is of the form

{'existing_param':'new_param'}

. However I'm getting an error similar to

kedro.pipeline.modular_pipeline.ModularPipelineError: Failed to map datasets and/or parameters: existing_param

. Digging up a bit I found the code that checks the existence of the parameters:

existing = {_strip_transcoding(ds) for ds in pipe.data_sets()}
    non_existent = (inputs | outputs | parameters) - existing
    if non_existent:
        raise ModularPipelineError(
            f"Failed to map datasets and/or parameters: "
            f"{', '.join(sorted(non_existent))}"
        )

What I don't understand is that

existing

only includes datasets, not params. What am I missing here 😅

If you're using namespaces it may be missing a prefix

So when the docs say

namespace (Optional[str]) – A prefix to give to all dataset names, except those explicitly named with the inputs/outputs arguments, and parameter references (params: and parameters).

I thought they meant parameters are not prefixed by namespace. In any case I tried adding some prefixes but still couldn't get it to work

@williamc which version of Kedro are you on? On 0.17.x Parameters are NOT namespaced, but after user feedback they are in 0.18.x

You can roll back the docs to a particular version at the bottom left https://kedro.readthedocs.io/en/0.17.7/kedro.pipeline.modular_pipeline.pipeline.html

Python Kedro PySpark : py4j.protocol.Py4JJavaError: An error occurred while calling None.org.apache.spark.api.java.JavaSparkContext https://stackoverflow.com/questions/71797449/python-kedro-pyspark-py4j-protocol-py4jjavaerror-an-error-occurred-while-call

I'm on 0.17.5

In which case this tutorial project should help! https://github.com/datajoely/modular-spaceflights

A module that I'm working with that uses

multiprocessing

gives me this complaint while being used within Kedro:

from pycbc.waveform import get_fd_waveform
  File "/Users/jordan/mambaforge/envs/gravitational_waves/lib/python3.8/site-packages/pycbc/__init__.py", line 150, in <module>
    multiprocessing.set_start_method('fork')
  File "/Users/jordan/mambaforge/envs/gravitational_waves/lib/python3.8/multiprocessing/context.py", line 243, in set_start_method
    raise RuntimeError('context has already been set')
RuntimeError: context has already been set

Does anyone have an idea how I can fix this? Cheers.

Hello - are you using parallel runner?

Hello! No, strictly speaking I'm not doing anything yet, it was actually just Kedro Viz that produced this error

In you settings.py could you change from shelve store to BaseSessionStore?

The error still seems to persist. To be clear, I've just uncommented the following lines from settings.py:

from kedro.framework.session.store import ShelveStore
SESSION_STORE_CLASS = ShelveStore

Is this what you meant?

The error can be reproduced easily in a Jupyter Notebook: Steps to reproduce: --> pip install pycbc into existing env which has kedro --> new Jupyter Notebook (ordinary, not Kedro)

%load_ext kedro.extras.extensions.ipython
from pycbc.waveform import get_fd_waveform

If the load order of the imports are switched, things work in a notebook, but doing anything from the command line results in the same error.

Yeah so with the CLI Kedro will import everything at once so with jupyter you're just dealying that conflict

So it fails even when doing this?

from kedro.framework.session.store import BaseSessionStore
SESSION_STORE_CLASS = BaseSessionStore

Yes, both settings yield the same error message, unfortunately!

Multiprocessing error

Hello, I am migrating an internal lib that we developped to

kedro==0.18.0

. We use

kedro.framework.session.get_current_session

to get the current session in order to either create a new session if it is None, or use it directly. This function was removed in

0.18.0

(with https://github.com/kedro-org/kedro/pull/1138) . What is the new way to find the current active session?