How to pass csv and dataframe between

[@AshutoshTripathi_AI]

How to install kubeflow pipeline locally: kzbin.info/www/bejne/goTZp3yqZ9JkgNE kubeflow basic pipeline creation Part 1: kzbin.info/www/bejne/q2LcqXmun8qSqtU

[@geetatripathi9335]

Good

[@AshutoshTripathi_AI]

🙏

[@mlworks]

Thank you @ashutosh. Great work.

[@AshutoshTripathi_AI]

Thank you.

[@pankajsuryavanshi8332]

Will you cover airflow dag if the when it will come?

[@AshutoshTripathi_AI]

Yes there is a plan. Will try to upload asap. Stay tuned. Thanks

[@ranjithkumarkarthic8696]

Hai @Ashuthosh, Thanks for your wonderful MLOPS playlist. I request you to make a End to End project implementation video utilizing DVC, MLflow, and Kubeflow such that everything can be done in a single flow. Also please create a video demonstrating how to use the Kubeflow pipeline to trigger the model retraining function by checking the model performance / by detecting the data drift. It will be very much helpful to understand the complete structure and the flow of MLops. I know you're going to upload your 2nd Video on DVC for data version after "Importance and Need of data versioning in Machine Learning: video. I'm waiting for that also. Please replay in comments if its possible. I'm waiting for your replay. Once again thank you so much.

[@AshutoshTripathi_AI]

Yes, exactly this is the plan, however it takes time to prepare the complete end to end flow. Just keep watching and you will find it soon. Also the next videos will be on DVC so stay tuned and enjoy learning.

[@ranjithkumarkarthic8696]

@@AshutoshTripathi_AI Thank you so much. I'm learning MLOPS through your videos. Fan of your teaching ❤️

[@venkateshsalla5820]

Hi Ashutosh, This video is very helpful for us. Also can you please make a video on visualizing in kubeflow components

[@AshutoshTripathi_AI]

Sure, will try to create one.

[@mlworks]

I have tried this in my project, while moving csv from one component to another, in your example you've used output_csv and are able to fetch using component.outputs['output_csv'], but I am unable to do that, I have used component.outputs['output_file'], where output_file is the file name which I have used, and it works if I give component.outputs['output']. Any idea why this happens?

[@AshutoshTripathi_AI]

Just check you have defined same name output_file in parameter of outputpath. It should work. If not just share the screenshot of the function and component I can check and reply.

[@mlworks]

@@AshutoshTripathi_AI unable to share the screenshot here.

[@AshutoshTripathi_AI]

@@mlworks you cand send on LinkedIn.

[@vishalwaghmare3130]

How can we use custom helper function inside the components?

[@venkateshsalla5820]

Hi @AshutoshTripathi_AI Can you please make a video on katlib and hyperparameter tuning

[@AshutoshTripathi_AI]

Ok, sure. Will upload soon.

[@keerthigavenkatesh3806]

how to handle the files locally, can we give local drive path for fetching input and storing output ?

[@sivakalahasti]

Hi Ashutosh, thanks for the demonstration, how do I figure out where my files are saving since I don't see them on the path I gave. Also, how do I save multiple .pkl files...ritenow it's only taknig the last ran .pkl file.

[@AshutoshTripathi_AI]

After every run, you can do dvc add and dvc push them git add git commit git tag git push. So your files will be pushed to dvc remote storage as binary large object. So if u go to the location you will see folders created with random names. Now to see the files u need to do git checkout tagname and dvc pull. So in ur working directory u will see the files.

[@siddharthtyagi1254]

sir please share this notebook

[@kanakorn]

Now the version run to KFP v2, will you update this tutorial ? Thanks.

[@RaushanKumar-ut2ke]

How we can import custom build python module in kubeflow pipeline for model training

[@RaushanKumar-ut2ke]

I have made one module for converting categorical columns to numerical, i have saved this class as .py file and using it, but i dont understand how to import this in kubeflow pipeline.

[@AshutoshTripathi_AI]

You can build individual components for those python modules and then use it as yaml inside the pipeline. You can refer to the kubeflow doc as I have not covered this part yet.

[@AshutoshTripathi_AI]

You can build an image for that python module and then import it. Refer the below url: stackoverflow.com/questions/73623659/is-it-possible-to-import-custom-source-files-into-kubeflow-components

[@RaushanKumar-ut2ke]

@@AshutoshTripathi_AI thank you, for the info, i will try to implement this way.

[@mukundvedanarayanan5316]

Hello Could you pls Share the jupyter notebook file for this.Would be a help.

[@girishai7058]

can i have the code please .pynb file

[@gichurumuthomi8578]

source code for this????