keyu-tian / SparK

[ICLR'23 Spotlight🔥] The first successful BERT/MAE-style pretraining on any convolutional network; Pytorch impl. of "Designing BERT for Convolutional Networks: Sparse and Hierarchical Masked Modeling"
https://arxiv.org/abs/2301.03580
MIT License
1.42k stars 82 forks source link

Is it possible to support the unsquare input #16

Closed zero0kiriyu closed 1 year ago

zero0kiriyu commented 1 year ago

Current implement only support the square input size, it would be great if this repo can support the unsquare input size

keyu-tian commented 1 year ago

Would be merged into #17