search

cuiziteng / Illumination-Adaptive-Transformer

🌕 [BMVC 2022] You Only Need 90K Parameters to Adapt Light: A Light Weight Transformer for Image Enhancement and Exposure Correction. SOTA for low light enhancement, 0.004 seconds try this for pre-processing.
Apache License 2.0
459 stars 43 forks source link

CUDA outof memory #28

Closed Mr-ind1fferent closed 1 year ago

Mr-ind1fferent commented 1 year ago

3060怎么推理跑不动会直接炸，用这个模型跑视频的

cuiziteng commented 1 year ago

多少显存的3060啊，输入尺寸多少？

你可以试试这个：torch.utils.checkpoint ，把这个加在模型(model.py)里，显存能减少好多。

Mr-ind1fferent commented 1 year ago

6gb

Mr-ind1fferent commented 1 year ago

6gb

是在推理过程中，非训练，显存不足

cuiziteng commented 1 year ago

哦哦，你把图像缩小，或者用这个torch.utils.checkpoint试试看呢。

Mr-ind1fferent commented 1 year ago

在哪里更改图片大小

cuiziteng commented 1 year ago

你加载的时候resize一下就好啦