Machine Learning: The Kernel appears to have died. Always when I am loading a large file

For anyone that is involved with Python and turicreate / CoreML.

I am using Jupyter notebooks for creating aML Model with TuriCreate.

The steps that I am following are the above.

I have download a .csv and .json (same file) from

The file has 51 MB size.

I have made an environment turienv From Anaconda Navigation and the following steps work perfect for smaller CSV / JSON files.

  1. source activate turienv
  2. pip install turicreate=5.0
  3. Jupiter notebook

----inside Notebook----

import turicreate as tc

wine_data = tc.SFrame.read_json('winemag-data-130k-v2.json', orient='records')

wine_data.head() <-- I see that everything is loaded properly

wine_model = tc.text_classifier.create(wine_data,'title',features=['description'])

PROGRESS: Creating a validation set from 5 percent of training data. This may take a while.
          You can set ``validation_set=None`` to disable validation tracking.

Logistic regression:
Number of examples          : 123481
Number of classes           : 113404
Number of feature columns   : 1
Number of unpacked features : 21030
Number of coefficients      : 2384978493
Starting L-BFGS
| Iteration | Passes   | Step size | Elapsed Time | Training Accuracy | Validation Accuracy |

And then in about 3-4 minutes I get error msg = The Kernel appears to have died.

Could anyone help?? I am new in Python and Jupyter is only env I have worked with. If there is an other env that I could run the same things with some guidance to have a more solid error msg that I could debug please let me know.

Edited: I am running the above on a 2018 MacBook Pro 16GB 512GB. I see on Activity Monitor Memory for python goes to 130GB and CPU 83%

Thanks in Advance

@audrey Can you please help with this when you get a chance? Thank you - much appreciated! :]

thanks @shogunkaramazov for the ping

hi scy, are you also grNorway on the apple/turicreate forum? have you tried TobyRoseman’s last suggestion?

Also, look at Matt Thomson’s tutorial using the Create ML text classifier — in Swift :slight_smile: — uses the same algorithms as TuriCreate, but with some features that help it run better on Macs.


This topic was automatically closed after 166 days. New replies are no longer allowed.