Closed xdobetter closed 1 year ago
Hello everyone,I have fixed the problem. My environment bellow: Windows10 with torch 1.12.0+cu113 & CUDA 11.3 on an RTX 3090 & tinycudann 1.7 I don't think the pytroch version is the cause of this issue as long as it matches the cuda version.I tried pytorch 1.11.0, 1.10.1, 1.12.1 , 1.12.0。
I fix the problem by change the folder freqencoder/gridencoder/raymarching/shencoder setup.py file
import os
from setuptools import setup
from torch.utils.cpp_extension import BuildExtension, CUDAExtension
_src_path = os.path.dirname(os.path.abspath(__file__))
nvcc_flags = [
'-O3', '-std=c++17', # set -std=c++17
'-U__CUDA_NO_HALF_OPERATORS__', '-U__CUDA_NO_HALF_CONVERSIONS__', '-U__CUDA_NO_HALF2_OPERATORS__',
]
I hope you find this information helpful.
my environment below:
Windows 10 with torch 1.11.0+cu113 & CUDA 11.3 on an RTX 3090 & tinycudann 1.7
I build torch-ngp according to "Install with pip", but I build extensions occurs errors.
The errors bellow:
others extensions are the same problem.
I look some issue #167 , and I try torch1.12.1,but it still don't work.
Could anyone give me some pieces of advice? Thanks!