fatal error: cuda_runtime_api.h: No such file or directory #15

imr555 · 2019-02-26T16:00:35Z

i set this
export CUDA_HOME="/usr/local/cuda/"
when i run "python setup.py install" i found this error.. how can i solve it ?
!! WARNING !!

warnings.warn(ABI_INCOMPATIBILITY_WARNING.format(compiler))
building 'warprnnt_pytorch.warp_rnnt' extension
gcc -pthread -B /home/imr555/miniconda3/envs/ariyan/compiler_compat -Wl,--sysroot=/ -Wsign-compare -DNDEBUG -g -fwrapv -O3 -Wall -Wstrict-prototypes -fPIC -I/home/imr555/Hasan/Project/RRNT/E2E-ASR-master/warp-transducer/include -I/home/imr555/miniconda3/envs/ariyan/lib/python3.6/site-packages/torch/lib/include -I/home/imr555/miniconda3/envs/ariyan/lib/python3.6/site-packages/torch/lib/include/torch/csrc/api/include -I/home/imr555/miniconda3/envs/ariyan/lib/python3.6/site-packages/torch/lib/include/TH -I/home/imr555/miniconda3/envs/ariyan/lib/python3.6/site-packages/torch/lib/include/THC -I/home/imr555/miniconda3/envs/ariyan/include/python3.6m -c src/binding.cpp -o build/temp.linux-x86_64-3.6/src/binding.o -std=c++11 -fPIC -DWARPRNNT_ENABLE_GPU -DTORCH_API_INCLUDE_EXTENSION_H -DTORCH_EXTENSION_NAME=warp_rnnt -D_GLIBCXX_USE_CXX11_ABI=0
cc1plus: warning: command line option ‘-Wstrict-prototypes’ is valid for C/ObjC but not for C++
In file included from /home/imr555/miniconda3/envs/ariyan/lib/python3.6/site-packages/torch/lib/include/THC/THCGeneral.h:12:0,
from /home/imr555/miniconda3/envs/ariyan/lib/python3.6/site-packages/torch/lib/include/THC/THC.h:4,
from src/binding.cpp:8:
/home/imr555/miniconda3/envs/ariyan/lib/python3.6/site-packages/torch/lib/include/ATen/cuda/CUDAStream.h:6:30: fatal error: cuda_runtime_api.h: No such file or directory
#include "cuda_runtime_api.h"
^
compilation terminated.
error: command 'gcc' failed with exit status 1

HawkAaron · 2019-02-27T00:12:47Z

before run "python setup.py install", you need to

load gcc 5
load cuda library, it should be the same as pytorch

the environments for reference:

export CUDA_HOME=$HOME/tools/cuda-9.0 # change to your path
export CUDA_TOOLKIT_ROOT_DIR=$CUDA_HOME
export LD_LIBRARY_PATH="$CUDA_HOME/extras/CUPTI/lib64:$LD_LIBRARY_PATH"
export LIBRARY_PATH=$CUDA_HOME/lib64:$LIBRARY_PATH
export LD_LIBRARY_PATH=$CUDA_HOME/lib64:$LD_LIBRARY_PATH
export CFLAGS="-I$CUDA_HOME/include $CFLAGS"

imr555 · 2019-02-27T09:49:04Z

It's working.. Thank you so much.

DemisEom · 2019-05-27T06:53:15Z

i have same problem...:(

what is that mean "load gcc5, load cuda library"??

HawkAaron · 2019-05-27T07:00:00Z

@shelling203 Compile the extension using gcc4.9 or higher, and load cuda library in your environment before compile with GPU support.

DemisEom · 2019-05-27T07:21:33Z

um... I'm sorry I do not understand.
My gcc version is 5.4.0....
Is it right that I have to run something else before I run "python setup.py install"?
or
I have to edit setup code?

HawkAaron · 2019-05-27T07:52:36Z

export CUDA_HOME=$HOME/tools/cuda-9.0 # change to your path
export CUDA_TOOLKIT_ROOT_DIR=$CUDA_HOME
export LD_LIBRARY_PATH="$CUDA_HOME/extras/CUPTI/lib64:$LD_LIBRARY_PATH"
export LIBRARY_PATH=$CUDA_HOME/lib64:$LIBRARY_PATH
export LD_LIBRARY_PATH=$CUDA_HOME/lib64:$LD_LIBRARY_PATH
export CFLAGS="-I$CUDA_HOME/include $CFLAGS"

@shelling203 Please set those environment variables before run python setup.py install.

DemisEom · 2019-05-27T07:59:48Z

Thanks for your answer :)
but I already set environment variables and same issue...

HawkAaron · 2019-05-27T10:51:47Z

@shelling203
Could you provide more information ? Such as CUDA version, machine, error info, ...

triple-tam · 2019-07-11T16:24:59Z

Setting the environment variables (ie vim ~/.bash_profile and source ~/.bash_profile) worked for me!
Note in case this is helpful for anyone else: Don't specify the cuda version in your path.
e.g. Let CUDA_HOME=/usr/local/cuda (or whatever your path is) instead of using CUDA_HOME=/usr/local/cuda-10.0

gccyxy · 2019-07-13T07:58:34Z

Thanks for your answer :)
but I already set environment variables and same issue...

I have the same questions with you, I tried it ,but it does not work

gccyxy · 2019-07-13T08:13:32Z

Setting the environment variables (ie vim ~/.bash_profile and source ~/.bash_profile) worked for me!
Note in case this is helpful for anyone else: Don't specify the cuda version in your path.
e.g. Let CUDA_HOME=/usr/local/cuda (or whatever your path is) instead of using CUDA_HOME=/usr/local/cuda-10.0

I have many cuda.h , How can I select the right CUDA_HOME？？

MY cuda.h is On where as followes:
/cm/shared/apps/hwloc/1.11.8/include/hwloc/cuda.h
/cm/shared/apps/tensorflow/1.7.0/lib/python2.7/site-packages/tensorflow/include/tensorflow/core/platform/cuda.h
/cm/shared/apps/openmpi/cuda/64/3.0.0/include/openmpi/opal/mca/hwloc/hwloc1117/hwloc/include/hwloc/cuda.h
/cm/shared/apps/hpcx/2.0.0/ompi-v3.0.x/include/openmpi/opal/mca/hwloc/hwloc1117/hwloc/include/hwloc/cuda.h
/cm/shared/apps/cuda91/toolkit/9.1.85/targets/x86_64-linux/include/cuda.h
/home/jupyter/anaconda3/pkgs/pytorch-1.0.1-cuda92py36h65efead_0/lib/python3.6/site-packages/torch/lib/include/torch/csrc/api/include/torch/cuda.h
/home/jupyter/anaconda3/pkgs/tensorflow-base-1.12.0-gpu_py36h8e0ae2d_0/lib/python3.6/site-packages/tensorflow/include/tensorflow/core/platform/cuda.h
/home/jupyter/anaconda3/pkgs/pytorch-1.0.1-cuda100py36he554f03_0/lib/python3.6/site-packages/torch/lib/include/torch/csrc/api/include/torch/cuda.h
/home/jupyter/anaconda3/pkgs/tensorflow-base-1.12.0-gpu_py36had579c0_0/lib/python3.6/site-packages/tensorflow/include/tensorflow/core/platform/cuda.h
/home/jupyter/anaconda3/lib/python3.6/site-packages/tensorflow/include/tensorflow/core/platform/cuda.h
/home/ai18/Gry/darknet/src/cuda.h
/home/ai18/Gry/darknet2/darknet/src/cuda.h
/home/ai18/Gry/naofinal/darknet_nao/src/cuda.h
/home/ai18/Gry/naofinal/darknet_mobilenet-master/src/cuda.h
/home/ai18/Gry/darknet_Mobile/src/cuda.h
/home/ai18/Gry/darknet_nao/src/cuda.h
/home/ai18/Gry/Gry/darknet/src/cuda.h
/home/ai18/Gry/darknet_mobilenet-master/src/cuda.h
/home/qr/miniconda3/pkgs/tensorflow-base-1.14.0-mkl_py37h7ce6ba3_0/lib/python3.7/site-packages/tensorflow/include/tensorflow/core/platform/cuda.h
/home/qr/miniconda3/lib/python3.7/site-packages/tensorflow/include/tensorflow/core/platform/cuda.h
/home/wbw/.local/lib/python2.7/site-packages/tensorflow/include/tensorflow/core/platform/cuda.h
/home/lwj/.pyenv/versions/3.5.2/lib/python3.5/site-packages/torch/lib/include/torch/csrc/api/include/torch/cuda.h
/home/lwj/.pyenv/versions/3.6.7/lib/python3.6/site-packages/tensorflow/include/tensorflow/core/p

HawkAaron · 2019-07-13T08:22:47Z

@gccyxy God knows how it goes.

Well, here is my suggestion:

install anaconda: https://www.anaconda.com/
create an env: `conda create -n <name you'd like> python=3.6
activate that env: . activate <name you'd like>
install whatever deep learning framework you want: conda install pytorch torchvision cudatoolkit=9.0 -c pytorch
go to the extension path, build everything from scratch

gccyxy · 2019-07-13T09:54:13Z

@gccyxy God knows how it goes.

Well, here is my suggestion:

install anaconda: https://www.anaconda.com/

create an env: `conda create -n <name you'd like> python=3.6

activate that env: . activate <name you'd like>

install whatever deep learning framework you want: conda install pytorch torchvision cudatoolkit=9.0 -c pytorch

go to the extension path, build everything from scratch

Thanks for you time !!! I am trying it now.
But My GCC version is 7.2 ,is it matter ?

HawkAaron · 2019-07-13T10:18:35Z

@gccyxy
It would be OK.

zhuhaozh · 2019-07-25T06:50:15Z

before run "python setup.py install", you need to

load gcc 5
load cuda library, it should be the same as pytorch

the environments for reference:

export CUDA_HOME=$HOME/tools/cuda-9.0 # change to your path
export CUDA_TOOLKIT_ROOT_DIR=$CUDA_HOME
export LD_LIBRARY_PATH="$CUDA_HOME/extras/CUPTI/lib64:$LD_LIBRARY_PATH"
export LIBRARY_PATH=$CUDA_HOME/lib64:$LIBRARY_PATH
export LD_LIBRARY_PATH=$CUDA_HOME/lib64:$LD_LIBRARY_PATH
export CFLAGS="-I$CUDA_HOME/include $CFLAGS"

Solved with this.
The GCC 5 is not necessary, gcc 4.8.5 in my environment, still works

anonymous-oracle · 2019-11-27T05:05:43Z

running install
running bdist_egg
running egg_info
writing warprnnt_pytorch.egg-info/PKG-INFO
writing dependency_links to warprnnt_pytorch.egg-info/dependency_links.txt
writing top-level names to warprnnt_pytorch.egg-info/top_level.txt
reading manifest file 'warprnnt_pytorch.egg-info/SOURCES.txt'
writing manifest file 'warprnnt_pytorch.egg-info/SOURCES.txt'
installing library code to build/bdist.linux-x86_64/egg
running install_lib
running build_py
running build_ext
building 'warprnnt_pytorch.warp_rnnt' extension
x86_64-linux-gnu-gcc -pthread -DNDEBUG -g -fwrapv -O2 -Wall -g -fstack-protector-strong -Wformat -Werror=format-security -Wdate-time -D_FORTIFY_SOURCE=2 -fPIC -I/home/suhas/warp-transducer/include -I/usr/local/lib/python3.6/dist-packages/torch/include -I/usr/local/lib/python3.6/dist-packages/torch/include/torch/csrc/api/include -I/usr/local/lib/python3.6/dist-packages/torch/include/TH -I/usr/local/lib/python3.6/dist-packages/torch/include/THC -I/usr/include/python3.6m -c src/binding.cpp -o build/temp.linux-x86_64-3.6/src/binding.o -std=c++11 -fPIC -DWARPRNNT_ENABLE_GPU -DTORCH_API_INCLUDE_EXTENSION_H -DTORCH_EXTENSION_NAME=warp_rnnt -D_GLIBCXX_USE_CXX11_ABI=1
In file included from /usr/local/lib/python3.6/dist-packages/torch/include/THC/THCGeneral.h:7:0,
from /usr/local/lib/python3.6/dist-packages/torch/include/THC/THC.h:4,
from src/binding.cpp:8:
/usr/local/lib/python3.6/dist-packages/torch/include/c10/cuda/CUDAStream.h:6:10: fatal error: cuda_runtime_api.h: No such file or directory
#include <cuda_runtime_api.h>
^~~~~~~~~~~~~~~~~~~~
compilation terminated.
error: command 'x86_64-linux-gnu-gcc' failed with exit status 1

I did all the configurations for the environment variables but couldn't get it to work

Running Ubuntu 18.04 LTS, CUDA-10.1, torch==1.4.0

aheba · 2020-03-27T15:09:22Z

It work fine for me with C++17

SaoYear · 2021-10-20T08:21:25Z

export CUDA_HOME=$HOME/tools/cuda-9.0 # change to your path
export CUDA_TOOLKIT_ROOT_DIR=$CUDA_HOME
export LD_LIBRARY_PATH="$CUDA_HOME/extras/CUPTI/lib64:$LD_LIBRARY_PATH"
export LIBRARY_PATH=$CUDA_HOME/lib64:$LIBRARY_PATH
export LD_LIBRARY_PATH=$CUDA_HOME/lib64:$LD_LIBRARY_PATH
export CFLAGS="-I$CUDA_HOME/include $CFLAGS"

@shelling203 Please set those environment variables before run python setup.py install.

YOUR ARE A LEGEND!!!!!!!!!!!

Lhx94As · 2022-04-19T13:22:57Z

before run "python setup.py install", you need to

1. load gcc 5

2. load cuda library, it should be the same as pytorch

the environments for reference:

export CUDA_HOME=$HOME/tools/cuda-9.0 # change to your path
export CUDA_TOOLKIT_ROOT_DIR=$CUDA_HOME
export LD_LIBRARY_PATH="$CUDA_HOME/extras/CUPTI/lib64:$LD_LIBRARY_PATH"
export LIBRARY_PATH=$CUDA_HOME/lib64:$LIBRARY_PATH
export LD_LIBRARY_PATH=$CUDA_HOME/lib64:$LD_LIBRARY_PATH
export CFLAGS="-I$CUDA_HOME/include $CFLAGS"

Works for me! Many thanks!

LiamLonergan · 2022-10-05T10:22:15Z

Thanks, this worked for me!

imr555 closed this as completed Feb 27, 2019

HawkAaron mentioned this issue Jun 28, 2019

unable to run test.py file #26

Closed

HawkAaron mentioned this issue Jul 12, 2019

GPU Test Failed #28

Closed

Peidong-Wang mentioned this issue Aug 22, 2019

environmental variables for warp-transducer during installation espnet/espnet#1123

Closed

qmpzzpmq mentioned this issue Aug 27, 2019

still pytorch binding GPU test error #39

Closed

shehrum mentioned this issue Nov 1, 2020

ESPnet installation issue - cuda_runtime_api.h: No such file or directory espnet/espnet#2615

Closed

This was referenced Dec 17, 2020

add warp transducer as submodule vincentqb/audio#5

Closed

add warp transducer pytorch/audio#1099

Closed

mrwyattii mentioned this issue Oct 31, 2022

"error: cuda_runtime_api.h: No such file or directory" microsoft/DeepSpeed-MII#75

Closed

mahmud30tibn mentioned this issue Feb 6, 2024

Setting up PyTorch plugin "upfirdn2d_plugin"... Failed! nv-tlabs/GET3D#18

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fatal error: cuda_runtime_api.h: No such file or directory #15

fatal error: cuda_runtime_api.h: No such file or directory #15

imr555 commented Feb 26, 2019

HawkAaron commented Feb 27, 2019

imr555 commented Feb 27, 2019

DemisEom commented May 27, 2019

HawkAaron commented May 27, 2019

DemisEom commented May 27, 2019

HawkAaron commented May 27, 2019

DemisEom commented May 27, 2019

HawkAaron commented May 27, 2019

triple-tam commented Jul 11, 2019

gccyxy commented Jul 13, 2019

gccyxy commented Jul 13, 2019

HawkAaron commented Jul 13, 2019

gccyxy commented Jul 13, 2019

HawkAaron commented Jul 13, 2019

zhuhaozh commented Jul 25, 2019

anonymous-oracle commented Nov 27, 2019

aheba commented Mar 27, 2020

SaoYear commented Oct 20, 2021

Lhx94As commented Apr 19, 2022

LiamLonergan commented Oct 5, 2022

fatal error: cuda_runtime_api.h: No such file or directory #15

fatal error: cuda_runtime_api.h: No such file or directory #15

Comments

imr555 commented Feb 26, 2019

HawkAaron commented Feb 27, 2019

imr555 commented Feb 27, 2019

DemisEom commented May 27, 2019

HawkAaron commented May 27, 2019

DemisEom commented May 27, 2019

HawkAaron commented May 27, 2019

DemisEom commented May 27, 2019

HawkAaron commented May 27, 2019

triple-tam commented Jul 11, 2019

gccyxy commented Jul 13, 2019

gccyxy commented Jul 13, 2019

HawkAaron commented Jul 13, 2019

gccyxy commented Jul 13, 2019

HawkAaron commented Jul 13, 2019

zhuhaozh commented Jul 25, 2019

anonymous-oracle commented Nov 27, 2019

aheba commented Mar 27, 2020

SaoYear commented Oct 20, 2021

Lhx94As commented Apr 19, 2022

LiamLonergan commented Oct 5, 2022