some questions about SER and RE task inference

[githublsk]

Hi wolfshow, I have a question on the belowing: I using Chinese document for layoutxlm downstream RE task following below steps: step1: finetuning a SER model with Chinese document; step2:finetuning a RE model on SER model trained by step 1; When I do inference, I should get SER result by SER model trained by step 1; then feed the SER results to RE model trained by step 2, is it right for this inference process? in this situation, to complete the RE task, I should use two models trained by step 1 and step 2?

[wolfshow]

@githublsk The RE model is often trained with keys and values with ground truth labels. So, basically, you need two models to finish the SER and RE tasks.

[githublsk]

@githublsk The RE model is often trained with keys and values with ground truth labels. So, basically, you need two models to finish the SER and RE tasks.

Thank you @wolfshow, when do inference, should I use SER model to get keys and values, and then use RE model to get the linking relation with keys and values outputed by SER model?