search

modelfoxdotdev / modelfox

ModelFox makes it easy to train, deploy, and monitor machine learning models.
Other
1.46k stars 63 forks source link

Training error when column to predict has more than 100 variants #120

Open joelchen opened 2 years ago

joelchen commented 2 years ago

When column to predict has more than 100 variants for multiclass classification, there is following error during training:

✅ Inferring train table columns. 6s
✅ Loading train table. 6s
✅ Shuffling. 0s 846ms
✅ Computing train stats. 10s
✅ Computing test stats. 2s
✅ Finalizing stats. 11s
error: invalid target column type
nitsky commented 2 years ago

Hi @joelchen the default settings assume that a column with more than 100 non-numeric unique values is a text column, not an enum column. You can force the CLI to treat your target column as an enum column using a config file.

joelchen commented 2 years ago

@nitsky Alright, the accuracy of 100 variants is low and I have not trained again with enum as target column in config file, but other users may encounter this issue, so I will leave it to your team to decide whether there is room for improvement.