[LayoutLM v2/v3] Could I use those models to infer the RE task on the FUNSD?

microsoft / unilm

Large-scale Self-supervised Pre-training Across Tasks, Languages, and Modalities

https://aka.ms/GeneralAI

MIT License

19.49k stars 2.48k forks source link

[LayoutLM v2/v3] Could I use those models to infer the RE task on the FUNSD? #1446

Open midsdsy opened 7 months ago

midsdsy commented 7 months ago

I would like to infer the LayoutLM v2/v3 models for the relation extraction task on the FUNSD dataset, so as the CORD dataset. The results reported in the paper are for the semantic entity recognition task. What should I do? Do I need to modify the model structure?

noorraghib12 commented 4 months ago

https://github.com/R0bk/transformers/blob/9c0e0ba9ccc0d32b795c2c0e0130931b92230292/src/transformers/models/layoutlmv2/outputs_layoutlmv2.py