Problem with Building

[DuckyBlender]

Log file: https://pastebin.com/uZCE9uE0
Won't compile successfully.
Using Windows 10 and Python 3.9.10
Just downloaded lastest CMake, CUDA and OptiX
GTX 1660 SUPER

[bolshoytoster]

@DuckyBlender does this help?

[quinwu]

I had the same problem, my GPU is 3090

System windows10
Python 3.9.10
CUDA 11.6.55
optiX 7.4.0

[bolshoytoster]

@quinwu does this help?

[pratikpadamwar]

I have set the TCNN_CUDA_ARCHITECTURES=86 so i was able to build with cmake but after build
when I run script/run.py am getting this following error
ERROR : Uncaught exception**: FullyFusedMLP: insufficient shared memory available on the GPU. Reduce n_neurons or use CutlassMLP (better compatibility but slower)**
is this error related to pytorch shared memory problem
I tried everything but it is not working
my GPU size 4096 MB
can anyone please help me out here

[Tom94]

You need to go into the .json config you're using (likely nerf/base.json if you're loading a NeRF dataset and not specifying a config) and replace FullyFusedMLP with CutlassMLP.

As the error message says, the GPU you have installed likely has too little shared memory for FullyFusedMLP. More info on shared memory here: https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#memory-hierarchy

[quinwu]

@quinwu does this help?

export TCNN_CUDA_ARCHITECTURES=86 is not a Windows command, I don't know how to change it

[bolshoytoster]

@quinwu it's an environment variable, how are you running the compiler?

[quinwu]

-- Performing Test HAS_MSVC_GL_LTCG - Success
-- Obtained target architecture from environment variable TCNN_CUDA_ARCHITECTURES=86
-- Targeting GPU architectures: 86
-- Configuring done
-- Generating done
-- Build files have been written to: C:/Users/yckj/Desktop/instant-ngp/build

I set CNN_CUDA_ARCHITECTURES=86 as environment variable, still have the same problem.

@DuckyBlender does this help?

[bolshoytoster]

@quinwu can you paste the ful error please.

[quinwu]

@quinwu can you paste the ful error please.

https://pastebin.com/hRVYhKZW

[bolshoytoster]

@quinwu did you delete the build/ directory and start over?

[quinwu]

@quinwu did you delete the build/ directory and start over?

yes, delete build/ and start over

[DuckyBlender]

@DuckyBlender does this help?

Still doesn't work
https://pastebin.com/7X3x9QZf

I made a work around by getting a compiled build from my friend.

[quinwu]

@DuckyBlender does this help?

Still doesn't work https://pastebin.com/7X3x9QZf

I made a work around by getting a compiled build from my friend.

compiled build tcnn?

[DuckyBlender]

@DuckyBlender does this help?

Still doesn't work https://pastebin.com/7X3x9QZf
I made a work around by getting a compiled build from my friend.

compiled build tcnn?

what does tcnn mean?

[quinwu]

@DuckyBlender does this help?

Still doesn't work https://pastebin.com/7X3x9QZf
I made a work around by getting a compiled build from my friend.

compiled build tcnn?

what does tcnn mean?

tiny-cuda-nn

[bolshoytoster]

@Hezhexi2002 the screenshot doesn't actually show any error, it just says that there were errors.

Could you paste the full error here (in triple backticks:
```
paste here
```

[Hezhexi2002]

OK,here is my CMakeOutput.log and CMakeError.log:
CMakeOutput.log
CMakeError.log

[bolshoytoster]

@Hezhexi2002 does this help?

[Hezhexi2002]

Sorry,the logs I have presented before seems not the true reason of the error,I try to generate the project in the VS again and increase verbosity of VS output,here is the output log:
ALL_BUILD.log

[Hezhexi2002]

I‘m so glad to tell you that I have successfully compiled the project after I install the CUDAv11.6,it seems that this is really due to the conflict between the nvcc of CUDAv11.3 and the MSVC of VS2019:

[DuckyBlender]

@DuckyBlender does this help?

Still doesn't work https://pastebin.com/7X3x9QZf

I made a work around by getting a compiled build from my friend.

compiled build tcnn?

what does tcnn mean?

tiny-cuda-nn

Idk man i just got the build folder from my friend and replaced mine.