when will the approach for the champion VCR datasets can be supported?

PaddlePaddle / ERNIE

Official implementations for various pre-training models of ERNIE-family, covering topics of Language Understanding & Generation, Multimodal Understanding & Generation, and beyond.

6.32k stars 1.28k forks source link

when will the approach for the champion VCR datasets can be supported? #515

Closed runzeer closed 3 years ago

yinweichong commented 4 years ago

scheduled to be supported in August, 2020.

yinweichong commented 4 years ago

sorry to notice that this work is postponed to Sep due to some emergency project.

TangDonnie commented 4 years ago

The pre-trained models of ERNIE-ViL and code for downstream VCR task have been released. https://github.com/PaddlePaddle/ERNIE/tree/repro/ernie-vil

HimariO commented 4 years ago

@TangDonnie will the code that generates image features LMDB files be released too?

TangDonnie commented 4 years ago

Our organization of visual features is modified from ViLBERT, thus we directly use the data from it. You can download them in https://github.com/jiasenlu/vilbert_beta/tree/master/data.

stale[bot] commented 3 years ago

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Feel free to reopen it. Thank you for your contributions.