Open entangledothers opened 2 years ago
ashawkey
commented
2 years ago HashEncoder (e.g., at /home/user/test/dreamfields-torch/encoding.py) and use the hash encoder of tiny-cuda-nn, by adding the --tcnn flag.
entangledothers
commented
2 years ago Of course!
´´´ OMP_NUM_THREADS=1 CUDA_VISIBLE_DEVICES=0 python main_nerf.py "cthulhu" --workspace trial --cuda_ray --fp16 --tcnn --gui
Namespace(H=800, W=800, aug_copy=8, bound=1, cuda_ray=True, dir_text=False, ff=False, fovy=90, fp16=True, gui=True, h=128, max_ray_batch=4096, max_spp=64, num_rays=4096, num_steps=128, radius=3, seed=0, tau_0=0.5, tau_1=0.8, tau_step=500, tcnn=True, test=False, text='cthulhu', upsample_steps=128, w=128, workspace='trial') Traceback (most recent call last): File "/home/user/anaconda3/envs/dreamfields/lib/python3.7/site-packages/torch/utils/cpp_extension.py", line 1723, in _run_ninja_build env=env) File "/home/user/anaconda3/envs/dreamfields/lib/python3.7/subprocess.py", line 512, in run output=stdout, stderr=stderr) subprocess.CalledProcessError: Command '['ninja', '-v']' returned non-zero exit status 1.
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "main_nerf.py", line 55, in
/home/user/dreamfields-torch/raymarching/src/raymarching.cu(271): warning: variable "d" was declared but never referenced detected during instantiation of "void kernel_composite_rays_train_forward(const scalar_t , const scalar_t , const scalar_t , const int , float, uint32_t, uint32_t, scalar_t , scalar_t ) [with scalar_t=float]" (444): here
/home/user/dreamfields-torch/raymarching/src/raymarching.cu(535): warning: variable "near" was declared but never referenced detected during instantiation of "void kernel_march_rays(uint32_t, uint32_t, const int , const scalar_t , const scalar_t , const scalar_t , float, uint32_t, const scalar_t , float, const scalar_t , const scalar_t , scalar_t , scalar_t , scalar_t , uint32_t) [with scalar_t=float]" (606): here
/usr/include/c++/10/chrono: In substitution of ‘template<class _Rep, class _Period> template~
Please submit a full bug report,
with preprocessed source if appropriate.
See file:///usr/share/doc/gcc-10/README.Bugs for instructions.
ninja: build stopped: subcommand failed.
´´´
ashawkey
commented
2 years ago @entangledothers It seems to be caused by the gcc version according to this issue. Could you try with a lower gcc version, such as gcc-9?
entangledothers
commented
2 years ago @entangledothers It seems to be caused by the gcc version according to this issue. Could you try with a lower gcc version, such as gcc-9?
Sadly, using gcc-9 (and even 8) made no difference, same issue as above.
ashawkey
commented
2 years ago Sorry for the late reply! A major updation has been pushed, you can try again to see if anything changes.
Have tried various env setups but get stuck with the following error when running this command:
OMP_NUM_THREADS=1 CUDA_VISIBLE_DEVICES=1 python main_nerf.py "cthulhu" --workspace trial --cuda_ray --fp16 --guiOnly thing changed in repo is having added verbose output to hashencoder. Also makes no difference with the tiny-cuda-nn installed or not.