search

jedyang97 / MTAG

Code for NAACL 2021 paper: MTAG: Modal-Temporal Attention Graph for Unaligned Human Multimodal Language Sequences
MIT License
43 stars 8 forks source link

measure metric #1

Closed bourne-3 closed 3 years ago

bourne-3 commented 3 years ago

Hi jed,your job,MTAG,is very excelent. I had some question of it. I had download the code and run it.I found that the test set in the code may not be used?And perform the measure in the validation after the training step? It is also applied in the past method like Mult?

jedyang97 commented 3 years ago

Hi @bourne-3, we are glad that you like our work!

The test set is not evaluated in the main training loop (only the validation set is evaluated during training). For evaluating on the test set, please use the eval_model() method provided in main.py. You will need to pass in a checkpoint that you want to use to evaluate.

The numbers reported in the paper, including those from MulT, are on the test set.