search

qianyuzqy / TransVOD_Lite

(TPAMI 2023) TransVOD:End-to-End Video Object Detection with Spatial-Temporal Transformers (implementations of TransVOD Lite).
Apache License 2.0
37 stars 6 forks source link

Batch Size > 1 caused error #16

Open fisal-ITS opened 1 year ago

fisal-ITS commented 1 year ago

I tried to use a batch size of more than 1 to train TransVOD Lite using a custom dataset. But I got the following error RuntimeError: The size of tensor a (28) must match the size of tensor b (14) at non-singleton dimension 0 . Is there a solution to this problem? Or maybe there is another solution so that I can speed up the training process a little.

fisal-ITS commented 1 year ago

help?

goku-krish10 commented 1 year ago

It's normal. always have batch size 1. The number of frames become the batch size, so the code is not adopted to have more than one batch size