airsplay / lxmert

PyTorch code for EMNLP 2019 paper "LXMERT: Learning Cross-Modality Encoder Representations from Transformers".
MIT License
923 stars 157 forks source link

why use tsv? #105

Closed Yucheng-Han closed 3 years ago

Yucheng-Han commented 3 years ago

In order to utilize existing code, I have to encode newly extracted features but found the process extremely slow by using array.tobytes().

Would there be any other choice rather than tsv? Or just because this function is inefficient?