search

NVlabs / PoseCNN-PyTorch

PyTorch implementation of the PoseCNN framework
Other
195 stars 49 forks source link

CUDA run out of memory #20

Open delonixsen opened 2 years ago

delonixsen commented 2 years ago

RuntimeError: CUDA out of memory. Tried to allocate 1.51 GiB (GPU 0; 3.82 GiB total capacity; 1.04 GiB already allocated; 74.38 MiB free; 1.04 GiB reserved in total by PyTorch)

Can someone help me with this problem? since I tried to reduce batch size, and nothing works

GPU : NVIDIA GeForce GTX 1650 4GB

Should I upgrade the Graphics Card?

gostensvig commented 2 years ago

I got the same error:

RuntimeError: CUDA out of memory. Tried to allocate 1.51 GiB (GPU 0; 3.82 GiB total capacity; 1.04 GiB already allocated; 99.75 MiB free; 1.04 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 error happened when I ran: ./experiments/scripts/demo.sh

GPU : NVIDIA Quadro T2000 with Max-Q

I'm using:

Full Traceback:

finished loading sdf ! 
Traceback (most recent call last):
  File "./tools/test_images.py", line 193, in <module>
    cfg.sdf_optimizer = sdf_optimizer(cfg.TEST.CLASSES[1:], sdf_files)
  File "/home/.../PoseCNN-PyTorch/tools/../lib/sdf/sdf_optimizer.py", line 73, in __init__
    self.sdf_torch = self.sdf_torch.cuda()
RuntimeError: CUDA out of memory. Tried to allocate 1.51 GiB (GPU 0; 3.82 GiB total capacity; 1.04 GiB already allocated; 99.75 MiB free; 1.04 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
lewamlyn commented 1 year ago

run ./experiments/scripts/demo.sh 0