Train recipe performance indicators

Hello... I'm wondering what the description of the performance indicators of the "train" and "train-curve" recipes is and how to interpret these indicators

For the "train" recipe, the output includes: LOSS TOK2VEC, LOSS NER, ENTS_F, ENTS_P, ENTS_R, SCORE
For the "train-curve" recipe, the output includes: Score, NER

I think SCORE is the model's F Score. I'm not sure about the rest.

I couldn't find a reference for these performance indicators in the documentation.

Thank you for your help in clarifying.

Hi @adixxov , to answer your question, yes the SCORE in the train-curve recipe corresponds to the F-score of the entities. For the train recipe, you can find a complete list of the output values in this page (cf. check the "Accuracy Evaluation" dropdown block in this link).

1 Like