search

Hlings / AsyFOD

(CVPR2023) The PyTorch implementation of the "AsyFOD: An Asymmetric Adaptation Paradigm for Few-Shot Domain Adaptive Object Detection".
MIT License
33 stars 3 forks source link

minimum gpu memory requird and runtime error while running YOLOs, YOLOm, YOLOl #4

Closed SumayyaInayat closed 10 months ago

SumayyaInayat commented 10 months ago

Hi, Hope you are doing well. Can you please tell me how much gpu ram is required to reproduce your results with YOLOx? And upon running YOLOs or YOLOm it gives dimension miss match error, Can you please help me with this?

RuntimeError: Given groups=1, weight of size [1280,1280,3,3], expected input[1,512,4,4] to have 1280 channels, but got 512 channels instead

Thanks!!!

Hlings commented 10 months ago

Hello, I ran all the experiments on V100 GPU, which has 32G memory :(

You can run results on YOLOs or YOLOm for comparisons. The key lies in the "Ranker" class here, which maps the feature map to embeddings for alignment. I specify the input dim as 1280, producing bugs when you use YOLOs or YOLOm. You can modify this hyperparameter or make it as an input argument.

Feel free to ask me if any other questions exist :)))

SumayyaInayat commented 10 months ago

Hi,

Oh ok so I was trying it on max 12 GB gpu with batch 2 but it gave Cuda out of memory error. I will make the required changes in Ranker class and then ask for any help if I want not able to solve it myself.

Thanks alott for helping!!