MUSK for VQA Tasks
#1
by
kaiko-ai-user
- opened
Dear authors,
I'm interested to evaluate MUSK for VQA tasks, but cannot find an example for text generation using MUSK on GitHub or Huggingface. Could you add such a use case to the model card or point me to a valid implementation using your model?
Many thanks!
Thank you for your interest in our work! Our approach does not involve a text generation module. Instead, we address the VQA problem by formulating it as a classification task on predefined answers and evaluating the accuracy (see Supplementary). You can refer to the codebase for the fine-tuning setup at https://github.com/microsoft/unilm/blob/master/beit/engine_for_finetuning.py.