I wanna ask about some definition of training metrics DIET Classifiers [image] what is t_loss, i_acc, and e_f1? i dont know what is t, i, and e before the underscore mean

t_loss is total loss I assume. i_loss - intent loss e_loss - entity loss t_loss = i_loss + e_loss + mask_loss i_acc - intent accuracy e_f1 - f1 score calculated on entity identification Ref: https://arxiv.org/pdf/2004.09936.pdf

I believe t is test, i is intent, e is entity. loss is the loss, acc is the accuracy, f1 is the F1 score. So it’s showing you the loss, intent extraction accuracy, and entity extraction F1 score. You can find detailed info on Rasa’s Algorithm Whiteboard .

Ask about Training Metrics

Rasa Open Source

hanifabd (Hanif Yuli Abdillah P) April 23, 2022, 10:00am 4

thank you very much @lis

Topic		Replies	Views
Get training metrics Rasa Open Source	0	439	February 12, 2019
Very high t_loss, but also with high m_acc and i_acc Rasa Open Source	14	2947	February 3, 2022
Output the training statistics Rasa Open Source	0	246	November 7, 2022
CRFEntityExtractor_report.json Rasa Open Source	2	369	February 13, 2020
Understanding more Tensorboard logs Rasa Open Source	0	188	August 11, 2023

Ask about Training Metrics

Related topics