Has anyone tried this implementation with NVIDIA RTX 4070 GPU ?

[PannagaS]

Hello, I am having a REALLY HARD TIME in implementing this project (NeuS2) on my local machine.
My specs:
NVIDIA RTX 4070, 32 GB ram , 8GB vram.
Ubuntu 22.04

I believe with the above configuration, I should be able to implement the project. However, I am constantly running into the following error when I execute the following two :

cmake . -B build
cmake --build build --config RelWithDebInfo -j

Can someone please help me out here? I would greatly appreciate any direction. I would like to know if someone tried this repo on a similar setup as mine. I think all this error is coming from the fact that the repo was designed and built on RTX 30xx GPU and I have RTX 40xx. I would greatly appreciate any kind of help.

Please feel free to reach out to via email - pannaga@umich.edu

Thank you very much.
Awaiting your kind support.

[erin47]

Hi @PannagaS , thank you for trying our work! We have tried to build our model on RTX 4090, and it runs smoothly. Could you print out and check your NVCC version? I think the issue might come from some conflicts from multiple cuda toolkits. You could refer to the issue here.

[PannagaS]

@erin47 thank you for responding. When I run cmake . -B build I get the following warnings :

(base) pannaga@legion7i:~/Documents/ROAHM_lab/NeuS2$ cmake . -B build
-- The C compiler identification is GNU 11.4.0
-- The CXX compiler identification is GNU 11.4.0
-- The CUDA compiler identification is NVIDIA 11.5.119
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Check for working C compiler: /usr/bin/cc - skipped
-- Detecting C compile features
-- Detecting C compile features - done
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Check for working CXX compiler: /usr/bin/c++ - skipped
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- Detecting CUDA compiler ABI info
-- Detecting CUDA compiler ABI info - done
-- Check for working CUDA compiler: /usr/bin/nvcc - skipped
-- Detecting CUDA compile features
-- Detecting CUDA compile features - done
-- No release type specified. Setting to 'Release'.
-- Targeting GPU architectures: 89
-- Found Vulkan: /usr/lib/x86_64-linux-gnu/libvulkan.so  
-- Looking for pthread.h
-- Looking for pthread.h - found
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD - Success
-- Found Threads: TRUE  
-- Using X11 for window creation
-- Found X11: /usr/include   
-- Looking for XOpenDisplay in /usr/lib/x86_64-linux-gnu/libX11.so;/usr/lib/x86_64-linux-gnu/libXext.so
-- Looking for XOpenDisplay in /usr/lib/x86_64-linux-gnu/libX11.so;/usr/lib/x86_64-linux-gnu/libXext.so - found
-- Looking for gethostbyname
-- Looking for gethostbyname - found
-- Looking for connect
-- Looking for connect - found
-- Looking for remove
-- Looking for remove - found
-- Looking for shmat
-- Looking for shmat - found
-- Looking for IceConnectionNumber in ICE
-- Looking for IceConnectionNumber in ICE - found
-- Found GLEW: /usr/include (found version "2.2.0") 
-- Found OpenMP_C: -fopenmp (found version "4.5") 
-- Found OpenMP_CXX: -fopenmp (found version "4.5") 
-- Found OpenMP: TRUE (found version "4.5")  
-- !!! Warning OptiX_INSTALL_DIR not set in environment. using default
-- OptiX_INSTALL_DIR value: /usr/local/NVIDIA-OptiX-SDK-7.3.0-linux64-x86_64
-- OptiX headers (optix.h and friends) not found.
CMake Warning at CMakeLists.txt:224 (message):
  OptiX was not found.  Neural graphics primitives will still compile and run
  correctly.  However, SDF training in 'raystab' and 'pathescape' modes will
  be significantly slower.


-- Found Python: /home/pannaga/anaconda3/bin/python3 (found suitable version "3.12.7", minimum required is "3.7") found components: Interpreter Development Development.Module Development.Embed 
-- pybind11 v2.7.1 
CMake Warning (dev) at /usr/share/cmake-3.22/Modules/CMakeDependentOption.cmake:84 (message):
  Policy CMP0127 is not set: cmake_dependent_option() supports full Condition
  Syntax.  Run "cmake --help-policy CMP0127" for policy details.  Use the
  cmake_policy command to set the policy and suppress this warning.
Call Stack (most recent call first):
  dependencies/pybind11/CMakeLists.txt:98 (cmake_dependent_option)
This warning is for project developers.  Use -Wno-dev to suppress it.

-- Performing Test HAS_FLTO
-- Performing Test HAS_FLTO - Success
-- Configuring done
-- Generating done
-- Build files have been written to: /home/pannaga/Documents/ROAHM_lab/NeuS2/build

But, when I run cmake --build build --config RelWithDebInfo -j, I get the following build error :

(base) pannaga@legion7i:~/Documents/ROAHM_lab/NeuS2$ cmake --build build --config RelWithDebInfo -j 
[  1%] Building C object dependencies/glfw/src/CMakeFiles/glfw_objects.dir/init.c.o
[  3%] Building C object dependencies/glfw/src/CMakeFiles/glfw_objects.dir/context.c.o
[  5%] Building CUDA object dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/network.cu.o
[  7%] Building CUDA object dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/encoding.cu.o
[  8%] Building C object dependencies/glfw/src/CMakeFiles/glfw_objects.dir/xkb_unicode.c.o
[ 14%] Building CUDA object dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/object.cu.o
[ 14%] Building C object dependencies/glfw/src/CMakeFiles/glfw_objects.dir/input.c.o
[ 14%] Building CUDA object dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/common.cu.o
[ 17%] Building CUDA object dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/common_device.cu.o
[ 17%] Building CUDA object dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/loss.cu.o
[ 23%] Building CUDA object dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/cpp_api.cu.o
[ 23%] Building C object dependencies/glfw/src/CMakeFiles/glfw_objects.dir/monitor.c.o
[ 23%] Building C object dependencies/glfw/src/CMakeFiles/glfw_objects.dir/window.c.o
[ 25%] Building C object dependencies/glfw/src/CMakeFiles/glfw_objects.dir/vulkan.c.o
[ 26%] Building C object dependencies/glfw/src/CMakeFiles/glfw_objects.dir/glx_context.c.o
[ 28%] Building C object dependencies/glfw/src/CMakeFiles/glfw_objects.dir/x11_monitor.c.o
[ 33%] Building CUDA object dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/reduce_sum.cu.o
[ 33%] Building CUDA object dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/fully_fused_mlp.cu.o
[ 35%] Building C object dependencies/glfw/src/CMakeFiles/glfw_objects.dir/x11_init.c.o
[ 33%] Building CUDA object dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/cutlass_mlp.cu.o
[ 37%] Building C object dependencies/glfw/src/CMakeFiles/glfw_objects.dir/egl_context.c.o
[ 39%] Building C object dependencies/glfw/src/CMakeFiles/glfw_objects.dir/osmesa_context.c.o
[ 41%] Building C object dependencies/glfw/src/CMakeFiles/glfw_objects.dir/posix_time.c.o
[ 42%] Building C object dependencies/glfw/src/CMakeFiles/glfw_objects.dir/x11_window.c.o
nvcc fatal   : Unsupported gpu architecture 'compute_89'
[ 46%] Building CUDA object dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/optimizer.cu.o
[ 46%] Building C object dependencies/glfw/src/CMakeFiles/glfw_objects.dir/posix_thread.c.o
nvcc fatal   : Unsupported gpu architecture 'compute_89'
gmake[2]: *** [dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/build.make:160: dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/network.cu.o] Error 1
gmake[2]: *** Waiting for unfinished jobs....
nvcc fatal   : Unsupported gpu architecture 'compute_89'
gmake[2]: *** [dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/build.make:174: dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/object.cu.o] Error 1
nvcc fatal   : Unsupported gpu architecture 'compute_89'
gmake[2]: *** [dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/build.make:132: dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/encoding.cu.o] Error 1
gmake[2]: *** [dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/build.make:76: dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/common.cu.o] Error 1
nvcc fatal   : Unsupported gpu architecture 'compute_89'
nvcc fatal   : Unsupported gpu architecture 'compute_89'
nvcc fatal   : Unsupported gpu architecture 'compute_89'
nvcc fatal   : Unsupported gpu architecture 'compute_89'
gmake[2]: *** [dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/build.make:90: dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/common_device.cu.o] Error 1
gmake[2]: *** [dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/build.make:104: dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/cpp_api.cu.o] Error 1
gmake[2]: *** [dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/build.make:118: dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/cutlass_mlp.cu.o] Error 1
nvcc fatal   : Unsupported gpu architecture 'compute_89'
gmake[2]: *** [dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/build.make:146: dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/loss.cu.o] Error 1
nvcc fatal   : Unsupported gpu architecture 'compute_89'
gmake[2]: *** [dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/build.make:202: dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/reduce_sum.cu.o] Error 1
gmake[2]: *** [dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/build.make:216: dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/fully_fused_mlp.cu.o] Error 1
nvcc fatal   : Unsupported gpu architecture 'compute_89'
gmake[2]: *** [dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/build.make:188: dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/optimizer.cu.o] Error 1
gmake[1]: *** [CMakeFiles/Makefile2:254: dependencies/neus2_tcnn/src/CMakeFiles/tiny-cuda-nn.dir/all] Error 2
gmake[1]: *** Waiting for unfinished jobs....
[ 48%] Building C object dependencies/glfw/src/CMakeFiles/glfw_objects.dir/linux_joystick.c.o
[ 48%] Built target glfw_objects
gmake: *** [Makefile:91: all] Error 2

nvcc version check:

(base) pannaga@legion7i:~/Documents/ROAHM_lab/NeuS2$ nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2024 NVIDIA Corporation
Built on Tue_Oct_29_23:50:19_PDT_2024
Cuda compilation tools, release 12.6, V12.6.85
Build cuda_12.6.r12.6/compiler.35059454_0

Can you please help me out here?

[erin47]

Hi @PannagaS , thank you for providing the information. The nvcc version of my 4090 is 11.8. Could you try to lower down your nvcc? The higher version might have some conflicts.

[PannagaS]

@erin47 I tried with cuda-toolkit version 11.8, but unfortunately still getting the same error. Then I read somewhere that I might have to upgrade my cuda-toolkit version; so I upgraded it to the latest 12.6 version, and yet the problem still persists. Nevertheless, I will try again. Meanwhile, can you tell me the output of ls /usr/local/ you are getting? I just want to confirm if having multiple cuda-toolkit versions won't be an issue.