Chapter 11 Notebook 5

junyi · June 5, 2021, 7:21am

There seems to be a problem when it comes to creating the lgbm model. I’ve tried debugging this myself but I have no idea why the type error is thrown since all parameters seem legit. Perhaps @Stefan can help with this? Thanks!

joseerasmosilva · June 5, 2021, 3:52pm

What environment are you using? I think this is a version problem.

junyi · June 6, 2021, 6:14am

I’m using the latest lightgbm version 3.2. I’m currently running it on my local jupyter notebook through anaconda.

joseerasmosilva · June 6, 2021, 5:09pm

Try to install the environment ml4t-base.yml or ml4t.yml.

Stefan · June 7, 2021, 4:59pm

This looks like an older version of the code; there was an error when creating train/validation subsets and the code how reads as:

        # iterate over folds
        for i, (train_idx, test_idx) in enumerate(cv.split(X=outcome_data)):
            
            # select train subset
            lgb_train = lgb_data.subset(used_indices=train_idx.tolist(),
                                       params=params).construct()

junyi · June 8, 2021, 12:01pm

Thanks Stefan, that helped.

rydeen7 · August 9, 2021, 6:25am

Hi, Stefan.
I fell same problem, but now this notebook runs with your sugested amendment.
I’m a bit surprized it takes about 30 hours to run through

anthonberg · March 8, 2023, 8:34pm

Hi, Is this notebook supposed to generate and save both the train and test dataset to the predictions.h5 file (later used in Chapter 11 Notebook 7)?

For me it only creates the test set.

Chapter 11 Notebook 7:

Thanks!