How to EASILY get the BEST Machine Learning Model with PyCaret (PyCarrot?) - AutoML in Python

[@GregHogg]

Take my courses at mlnow.ai/!

[@Value_Pilgrim]

Coming here after a long time. Found this very, very interesting. Thanks.

[@GregHogg]

Welcome back!

[@elizabethmorales9469]

I think this channel is really cool thank you for your knowledge!

[@GregHogg]

I really appreciate that Elizabeth!

[@mikekertser5384]

Thank you. It is very nice indeed, but seems to perform slowly on a large datasets, at least without GPU support. Saving the blended finalized models may result in surprisingly large files.

[@GregHogg]

Hey Mike! Yeah, luckily they said it should be much faster with a GPU. On a cluster I'm sure it'll be good, although personal computers wouldn't have access to that

[@superfreiheit1]

Great video and a great packet. It would be great to have a build in hyperparameter tuning in carets

[@mkvalor]

Great content. My immediate question is whether this framework works well for time series data which cannot be randomly shuffled, since it seems to do a lot of folding and cross validation on its own. Do you know if it has good options to keep the data ordered while still maintaining the ability to choose the best models?

[@FaroukYahaya-xi4ri]

what if i dont want to use the best model. but i want to select a particular model. so how do we change the print(best) to that specific model and finalize?

[@pramishprakash]

Sooper cool... Thanks alot sir

[@hivinline5184]

i'm sorry, i am new at this. i already use my own dataset for this, but it's doesn't work. even tho i already do many processing with that dataset. is it some thing wrong with the dataset or the pycaret can only do its own dataset? (sorry english is not my first language)

[@TS-fp5xg]

well explained ,thank you so much

[@GregHogg]

Great to hear, you're very welcome!

[@rgoudie]

Great stuff, Greg. Can you make a video on how to use PyCaret with time series? I am particularly interested in time series classification. Thanks.

[@GregHogg]

Hey Ray! Yeah that's on my to-do list for sure.

[@jacktrainer4387]

pyCaret took 2 hours and 53 minutes to train lgbm, gbm, and a dummy model on Google Colab Pro Plus (yes, the big guns). I could have done the models with pen and paper faster. You should have mentioned this in your video.

[@librae-5664]

thanks a lot!! one questions, how to edit n_estimators hyperparameters of base model?

[@GregHogg]

You can use the create_model function. Check the docs :)

[@imveryhungry112]

is there one for finding best nueral network model?

[@SP-db6sh]

Thank you. Plz Video on training DL model using Pycaret GPU

[@nataliaresende1121]

Do you know any similar approach to get the best neural network model?

[@GregHogg]

Yes, I made something similar for tensorflow recently. Also, pycaret will consider neural networks since basic feed forward nets are included in sklearn.

[@arsheyajain7055]

Ooh amazing!

[@GregHogg]

Thank you!

[@Mike-vj8do]

Hey bro, I really love the video. I'm wondering whether you think this automated ML engineers' jobs? Thank you very much.

[@GregHogg]

It definitely helps ML engineers. And glad to hear it!!

[@tcrnilo]

compare_models() gives nothing, does anyone have an idea?

[@GregHogg]

Version might have changed...?

[@omario.tntech]

another 🔥

[@omario.tntech]

Can I use this with CNN?

[@GregHogg]

Glad to hear it, and there's definitely some automated CNN libraries

[@motaseamyousef7668]

OMG! better than grid search 😂😂😂

[@GregHogg]

Haha yup!

[@patricionovellino6793]

Hi greg. Amazing video! I am going to try it with my actual and first project. What environment do you recommend(colab or jupyter)? Given the fact my dataset has 780k rows

[@GregHogg]

For 780k rows, you might need a cluster. Colab definitely won't work