How to test model (nighly) accuracy on eval-data.json?

,
1

I use data-to-spacy with split = 0.2 to create train-data.json and eval-data.json. Then I convert train-data.json to train-data.spacy and train using the new spacy pipeline. So how can I test this model on the eval-data.json?

2

Hi! I hope I understand your question correctly – but it sounds like the easiest solution would be to use spacy evaluate with your converted eval data? See here: https://nightly.spacy.io/api/cli#evaluate

3

Thanks. Can I convert eval-data.json to eval-data.jsonl so that I can test the accuracy in python, but not using the command line?