test error after installation: too many resources requested for launch

[lllabmaster]

hi, when i ran the test.py after installation following:

cd models/dino/ops
python setup.py build install
# unit test (should see all checking is True)
python test.py
cd ../../..

it raised that:
error in ms_deformable_col2im_cuda: too many resources requested for launch
error in ms_deformable_col2im_cuda: too many resources requested for launch
error in ms_deformable_col2im_cuda: too many resources requested for launch
File "/data/jilichuan/database/code/DINO-UW/models/dino/ops/test.py", line 76, in check_gradient_numerical
gradok = gradcheck(func, (value.double(), shapes, level_start_index, sampling_locations.double(), attention_weights.double(), im2col_step))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/data/anaconda3/envs/dino/lib/python3.11/site-packages/torch/autograd/gradcheck.py", line 2052, in gradcheck
return _gradcheck_helper(**args)
^^^^^^^^^^^^^^^^^^^^^^^^^
File "/data/anaconda3/envs/dino/lib/python3.11/site-packages/torch/autograd/gradcheck.py", line 2081, in _gradcheck_helper
_gradcheck_real_imag(
File "/data/anaconda3/envs/dino/lib/python3.11/site-packages/torch/autograd/gradcheck.py", line 1491, in _gradcheck_real_imag
gradcheck_fn(
File "/data/anaconda3/envs/dino/lib/python3.11/site-packages/torch/autograd/gradcheck.py", line 1632, in _slow_gradcheck
raise GradcheckError(
torch.autograd.gradcheck.GradcheckError: Jacobian mismatch for output 0 with respect to input 0,

did anyone encounter this error?

Then i skiped this error and ran the fine-tuning process, its ok and can ran well.

[lllabmaster]

my environment:
cuda: 11.8
torch: 2.5.0+cu118
os: ubuntu
gpu: v100-32G

[lllabmaster]

i fixed this error following chengdazhi/Deformable-Convolution-V2-PyTorch#15 (comment)

it seemed like my gpu is a litte out-date? 😄