You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I try to train rhino with the config 'rhino-4scale_r50_2xb2-12e_dotav15.py' on RTX 4090(24 G)×2,batch size is 2.But the following error message has appeared:
RuntimeError: CUDA out of memory. Tried to allocate 1.03 GiB (GPU 0; 23.65 GiB total capacity; 19.63 GiB already allocated; 735.00 MiB free; 20.85 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF
The text was updated successfully, but these errors were encountered:
@kouyuanbo I am sorry for the late response. Can you recheck your settings?
In my case, the same setting 'rhino-4scale_r50_2xb2-12e_dotav15.py' with batch_size=2 requires 8300~1570 MB.
Note that 'the denoising training' from DINO can result in higher memory usage when there are a lot of GT objects.
I try to train rhino with the config
'rhino-4scale_r50_2xb2-12e_dotav15.py'
on RTX 4090(24 G)×2,batch size is 2.But the following error message has appeared:The text was updated successfully, but these errors were encountered: