Thank you for your excellent work and the detailed environment setup process.
However, when using LE3D Editor to rendering the defocus effect, I encountered the following error during execution:
Traceback (most recent call last):
File "/workspaces/l3d/LE3D/basicgs/view.py", line 68, in start_viewer
viewer.update()
File "/usr/local/conda/envs/basicgs/lib/python3.10/site-packages/torch/autograd/grad_mode.py", line 27, in decorate_context
return func(*args, **kwargs)
File "/workspaces/l3d/LE3D/basicgs/viewers/base_viewer.py", line 338, in update
rendered_image, render_time, _ = self.get_current_image(camera_pack)
File "/usr/local/conda/envs/basicgs/lib/python3.10/site-packages/torch/autograd/grad_mode.py", line 27, in decorate_context
return func(*args, **kwargs)
File "/workspaces/l3d/LE3D/basicgs/viewers/le3d_viewer.py", line 53, in get_current_image
rendered_pkg = apply_defocus(rendered_image, hists, target_bin, 32,
File "/workspaces/l3d/LE3D/basicgs/utils/raw_util.py", line 565, in apply_defocus
render = _apply_defocus(mpi_rep, hists, kernel_size, translation)
File "/workspaces/l3d/LE3D/basicgs/utils/raw_util.py", line 534, in _apply_defocus
kernels = _get_gaussian_kernels_freq(kernel_sizes, H, W)
File "/workspaces/l3d/LE3D/basicgs/utils/raw_util.py", line 512, in _get_gaussian_kernels_freq
kernel_freq = torch.fft.fft2(kernel, s=(H_prime, W_prime))
RuntimeError: cuFFT error: CUFFT_INTERNAL_ERROR
2025-03-01 17:49:26,228 ERROR: Restarting viewer...
According to online sources, this issue is caused by compatibility problems between CUDA and PyTorch.
Some suggest upgrading to CUDA 11.8 and PyTorch 2.0, but I noticed that in your setup process, you used
conda install pytorch==1.12.1 torchvision==0.13.1 cudatoolkit=11.6 -c pytorch -c conda-forge -y.
When I upgrade, there are other errors. My environment is WSL virtual machine + Docker (CUDA: 11.8.0-cudnn8, Python 3.10).
Could you help me with this error?
Looking forward to your reply.
Many thanks.
Thank you for your excellent work and the detailed environment setup process.
However, when using LE3D Editor to rendering the defocus effect, I encountered the following error during execution:
According to online sources, this issue is caused by compatibility problems between CUDA and PyTorch.
Some suggest upgrading to CUDA 11.8 and PyTorch 2.0, but I noticed that in your setup process, you used
When I upgrade, there are other errors. My environment is WSL virtual machine + Docker (CUDA: 11.8.0-cudnn8, Python 3.10).
Could you help me with this error?
Looking forward to your reply.
Many thanks.