CUDA out of memory - Githubissues

Hello, @Paper99, @LGYoung, @NK-CS-ZZL I am trying to launch the evaluation script using CUDA GPU.

I ensured my PC has GPU enabled

>>> import torch

>>> torch.cuda.is_available()
True

>>> torch.cuda.device_count()
1

>>> torch.cuda.current_device()
0

>>> torch.cuda.device(0)
<torch.cuda.device at 0x7efce0b03be0>

>>> torch.cuda.get_device_name(0)
'GeForce GTX 950M'

I am using Windows 11

When I run the evaluation script, the following error is received

(ttt) F:\training\E2FGVI>python evaluate.py --model e2fgvi --dataset davis --data_root datasets/ --ckpt release_model/E2FGVI-CVPR22.pth
C:\Users\postu\miniconda3\envs\ttt\lib\site-packages\mmcv\__init__.py:21: UserWarning: On January 1, 2023, MMCV will release v2.0.0, in which it will remove components related to the training process and add a data transformation module. In addition, it will rename the package names mmcv to mmcv-lite and mmcv-full to mmcv. See https://github.com/open-mmlab/mmcv/blob/master/docs/en/compatibility.md for more details.
  'On January 1, 2023, MMCV will release v2.0.0, in which it will remove '
load pretrained SPyNet...
load checkpoint from http path: https://download.openmmlab.com/mmediting/restorers/basicvsr/spynet_20210409-c6c1bd09.pthLoading from: release_model/E2FGVI-CVPR22.pth
Start evaluation...
[Loading I3D model from ./release_model/i3d_rgb_imagenet.pt for FID score ..]
Traceback (most recent call last):
  File "evaluate.py", line 176, in <module>
    main_worker(args)
  File "evaluate.py", line 92, in main_worker
    pred_img, _ = model(masked_frames, len(neighbor_ids))
  File "C:\Users\postu\miniconda3\envs\ttt\lib\site-packages\torch\nn\modules\module.py", line 550, in __call__
    result = self.forward(*input, **kwargs)
  File "F:\training\E2FGVI\model\e2fgvi.py", line 255, in forward
    trans_feat = self.transformer(trans_feat)
  File "C:\Users\postu\miniconda3\envs\ttt\lib\site-packages\torch\nn\modules\module.py", line 550, in __call__
    result = self.forward(*input, **kwargs)
  File "C:\Users\postu\miniconda3\envs\ttt\lib\site-packages\torch\nn\modules\container.py", line 100, in forward
    input = module(input)
  File "C:\Users\postu\miniconda3\envs\ttt\lib\site-packages\torch\nn\modules\module.py", line 550, in __call__
    result = self.forward(*input, **kwargs)
  File "F:\training\E2FGVI\model\modules\tfocal_transformer.py", line 523, in forward
    mask_all=x_window_masks_all)  # nW*B, T*window_size*window_size, C
  File "C:\Users\postu\miniconda3\envs\ttt\lib\site-packages\torch\nn\modules\module.py", line 550, in __call__
    result = self.forward(*input, **kwargs)
  File "F:\training\E2FGVI\model\modules\tfocal_transformer.py", line 394, in forward
    attn = self.softmax(attn)
  File "C:\Users\postu\miniconda3\envs\ttt\lib\site-packages\torch\nn\modules\module.py", line 550, in __call__
    result = self.forward(*input, **kwargs)
  File "C:\Users\postu\miniconda3\envs\ttt\lib\site-packages\torch\nn\modules\activation.py", line 1044, in forward
    return F.softmax(input, self.dim, _stacklevel=5)
  File "C:\Users\postu\miniconda3\envs\ttt\lib\site-packages\torch\nn\functional.py", line 1442, in softmax
    ret = input.softmax(dim)
RuntimeError: CUDA out of memory. Tried to allocate 668.00 MiB (GPU 0; 4.00 GiB total capacity; 2.36 GiB already allocated; 0 bytes free; 3.05 GiB reserved in total by PyTorch)

Is there something that can be done here? Or is my PC hardware is too weak to launch the evaluation process?

MCG-NKU / E2FGVI

CUDA out of memory #66