Open shanhaidexiamo opened 4 years ago
I'm sure you must have changed something, e.g. the makefile, or your distribution is not up to date.
On Tue, Sep 1, 2020 at 10:23 AM shanhaidexiamo notifications@github.com wrote:
When I try to make depend in src, there always sth wrong , such as "feature-online-cmvn-cuda.cu:19:47: fatal error: cudafeat/feature-online-cmvn-cuda.h: No such file or directory". And I find that these files are internal files in kaldi, they failed when they include each other. On the other hand, it does not report errors when I change include "cudafeat/feature-online-cmvn-cuda.h" to an absolute path, but that's not a good idea. Can someone tell me what the problem is and how to solve it? Thank you !
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/kaldi-asr/kaldi/issues/4252, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAZFLO23UK6H5SSXR25BUEDSDRLJDANCNFSM4QRFNTTQ .
ok,thank you. actually I have download many times, and I did not change the makefile. maybe the env variable have sth wrong…
---Original--- From: "Daniel Povey"<notifications@github.com> Date: Tue, Sep 1, 2020 11:31 AM To: "kaldi-asr/kaldi"<kaldi@noreply.github.com>; Cc: "shanhaidexiamo"<946712859@qq.com>;"Author"<author@noreply.github.com>; Subject: Re: [kaldi-asr/kaldi] make depend (#4252)
I'm sure you must have changed something, e.g. the makefile, or your distribution is not up to date.
On Tue, Sep 1, 2020 at 10:23 AM shanhaidexiamo <notifications@github.com> wrote:
> When I try to make depend in src, there always sth wrong , such as > "feature-online-cmvn-cuda.cu:19:47: fatal error: > cudafeat/feature-online-cmvn-cuda.h: No such file or directory". And I find > that these files are internal files in kaldi, they failed when they include > each other. On the other hand, it does not report errors when I change > include "cudafeat/feature-online-cmvn-cuda.h" to an absolute path, but > that's not a good idea. Can someone tell me what the problem is and how to > solve it? Thank you ! > > — > You are receiving this because you are subscribed to this thread. > Reply to this email directly, view it on GitHub > <https://github.com/kaldi-asr/kaldi/issues/4252>, or unsubscribe > <https://github.com/notifications/unsubscribe-auth/AAZFLO23UK6H5SSXR25BUEDSDRLJDANCNFSM4QRFNTTQ> > . >
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.
I meet a similar problem when make depend and the error is:
online-ivector-feature-cuda-kernels.cu:19:58: fatal error: cudafeat/online-ivector-feature-cuda-kernels.h: No such file or directory
include "cudafeat/online-ivector-feature-cuda-kernels.h"
Besides, there is no any error when I back to former version.
I meet a similar problem when make depend and the error is:
online-ivector-feature-cuda-kernels.cu:19:58: fatal error: cudafeat/online-ivector-feature-cuda-kernels.h: No such file or directory
include "cudafeat/online-ivector-feature-cuda-kernels.h"
Besides, there is no any error when I back to former version.
I am sure this is a series errors, so your problem is same as me. Besides, I find this problem in a new server, when I turn back to old server, everything is ok
maybe try "make clean" before make depend, just to make sure... Plus, we saw several issues with virtual machines using SMB volumes (window shares) y,
On Tue, Sep 1, 2020 at 7:50 AM shanhaidexiamo notifications@github.com wrote:
I meet a similar problem when make depend and the error is:
online-ivector-feature-cuda-kernels.cu:19:58: fatal error: cudafeat/online-ivector-feature-cuda-kernels.h: No such file or directory
include "cudafeat/online-ivector-feature-cuda-kernels.h"
Besides, there is no any error when I back to former version.
I am sure this is a series errors, so your problem is same as me. Besides, I find this problem in a new server, when I turn back to old server, everything is ok
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/kaldi-asr/kaldi/issues/4252#issuecomment-684375857, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACUKYXYJ6ZB437A5IDLHEE3SDSDTHANCNFSM4QRFNTTQ .
Downloaded from where? How?
On Tue, Sep 1, 2020 at 2:57 PM jtrmal notifications@github.com wrote:
maybe try "make clean" before make depend, just to make sure... Plus, we saw several issues with virtual machines using SMB volumes (window shares) y,
On Tue, Sep 1, 2020 at 7:50 AM shanhaidexiamo notifications@github.com wrote:
I meet a similar problem when make depend and the error is:
online-ivector-feature-cuda-kernels.cu:19:58: fatal error: cudafeat/online-ivector-feature-cuda-kernels.h: No such file or directory
include "cudafeat/online-ivector-feature-cuda-kernels.h"
Besides, there is no any error when I back to former version.
I am sure this is a series errors, so your problem is same as me. Besides, I find this problem in a new server, when I turn back to old server, everything is ok
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/kaldi-asr/kaldi/issues/4252#issuecomment-684375857, or unsubscribe < https://github.com/notifications/unsubscribe-auth/ACUKYXYJ6ZB437A5IDLHEE3SDSDTHANCNFSM4QRFNTTQ
.
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/kaldi-asr/kaldi/issues/4252#issuecomment-684491543, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAZFLO4M3RMDG26YYAT22B3SDSLPLANCNFSM4QRFNTTQ .
Downloaded from where? How? … On Tue, Sep 1, 2020 at 2:57 PM jtrmal @.> wrote: maybe try "make clean" before make depend, just to make sure... Plus, we saw several issues with virtual machines using SMB volumes (window shares) y, On Tue, Sep 1, 2020 at 7:50 AM shanhaidexiamo @.> wrote: > I meet a similar problem when make depend and the error is: > > online-ivector-feature-cuda-kernels.cu:19:58: fatal error: > cudafeat/online-ivector-feature-cuda-kernels.h: No such file or directory > #include "cudafeat/online-ivector-feature-cuda-kernels.h" > > Besides, there is no any error when I back to former version. > > I am sure this is a series errors, so your problem is same as me. Besides, > I find this problem in a new server, when I turn back to old server, > everything is ok > > — > You are receiving this because you are subscribed to this thread. > Reply to this email directly, view it on GitHub > <#4252 (comment)>, > or unsubscribe > < https://github.com/notifications/unsubscribe-auth/ACUKYXYJ6ZB437A5IDLHEE3SDSDTHANCNFSM4QRFNTTQ > > . > — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#4252 (comment)>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAZFLO4M3RMDG26YYAT22B3SDSLPLANCNFSM4QRFNTTQ .
just use git clone https://github.com/kaldi-asr/kaldi.git and then compile tools ,and then src. but when make depend in src, there have a problem : `#$ SPACE=
online-ivector-feature-cuda-kernels.cu:19:58: fatal error: cudafeat/online-ivector-feature-cuda-kernels.h: No such file or directory compilation terminated.
../makefiles/default_rules.mk:149: recipe for target 'depend' failed make[1]: [depend] Error 1 make[1]: Leaving directory '/data/jiangyv/tp/kaldi/src/cudafeat' Makefile:127: recipe for target 'cudafeat/depend' failed make: [cudafeat/depend] Error 2 `
Probably because the current directory "." is not on the list of includes. Perhaps you did not run configure. Did you look at the INSTALL instructions?
On Tue, Sep 1, 2020 at 4:00 PM shanhaidexiamo notifications@github.com wrote:
Downloaded from where? How? … <#m3399741452031424692> On Tue, Sep 1, 2020 at 2:57 PM jtrmal @.> wrote: maybe try "make clean" before make depend, just to make sure... Plus, we saw several issues with virtual machines using SMB volumes (window shares) y, On Tue, Sep 1, 2020 at 7:50 AM shanhaidexiamo @.> wrote: > I meet a similar problem when make depend and the error is: > > online-ivector-feature-cuda-kernels.cu:19:58: fatal error: > cudafeat/online-ivector-feature-cuda-kernels.h: No such file or directory >
include "cudafeat/online-ivector-feature-cuda-kernels.h" > > Besides,
there is no any error when I back to former version. > > I am sure this is a series errors, so your problem is same as me. Besides, > I find this problem in a new server, when I turn back to old server, > everything is ok
— > You are receiving this because you are subscribed to this thread. > Reply to this email directly, view it on GitHub > <#4252 (comment) https://github.com/kaldi-asr/kaldi/issues/4252#issuecomment-684375857>, or unsubscribe > < https://github.com/notifications/unsubscribe-auth/ACUKYXYJ6ZB437A5IDLHEE3SDSDTHANCNFSM4QRFNTTQ . > — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#4252 (comment) https://github.com/kaldi-asr/kaldi/issues/4252#issuecomment-684491543>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAZFLO4M3RMDG26YYAT22B3SDSLPLANCNFSM4QRFNTTQ .
just use git clone https://github.com/kaldi-asr/kaldi.git https://github.com/kaldi-asr/kaldi.git and then compile tools ,and then src. but when make depend in src, there have a problem : `#$ SPACE=
$ CUDART=cudart
$ HERE=/usr/local/cuda/bin
$ THERE=/usr/local/cuda/bin
$ TARGET_SIZE=
$ TARGET_DIR=
$ TARGET_SIZE=64
$ TOP=/usr/local/cuda/bin/..
$ NVVMIR_LIBRARY_DIR=/usr/local/cuda/bin/../nvvm/libdevice
$ LD_LIBRARY_PATH=/usr/local/cuda/bin/../lib:
$
PATH=/usr/local/cuda/bin/../nvvm/bin:/usr/local/cuda/bin:/data/jiangyv/anaconda3/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games
$ INCLUDES="-I/usr/local/cuda/bin/..//include"
$ LIBRARIES= "-L/usr/local/cuda/bin/..//lib64/stubs"
"-L/usr/local/cuda/bin/..//lib64"
$ CUDAFE_FLAGS=
$ PTXAS_FLAGS=
$ g++ -std=c++14 -D__CUDA_ARCH__=300 -E -x c++
-DCUDA_DOUBLE_MATH_FUNCTIONS -DCUDACC -DNVCC -I"/usr/local/cuda/include" -I"/data/jiangyv/tp/kaldi/tools/cub-1.8.0" "-I/usr/local/cuda/bin/..//include" -D "HAVE_CUDA" -D "KALDI_DOUBLEPRECISION=0" -D "CUDA_API_PER_THREAD_DEFAULT_STREAM" -DCUDACC_VER_MAJOR=10 -DCUDACC_VER_MINOR=0 -DCUDACC_VER_BUILD=130 -include "cuda_runtime.h" -m64 " online-ivector-feature-cuda-kernels.cu" > "/tmp/tmpxft_00003be6_00000000-4_online-ivector-feature-cuda-kernels.cpp1.ii" online-ivector-feature-cuda-kernels.cu:19:58: fatal error: cudafeat/online-ivector-feature-cuda-kernels.h: No such file or directory compilation terminated. --error 0x1 --
../makefiles/default_rules.mk:149: recipe for target 'depend' failed make[1]: [depend] Error 1 make[1]: Leaving directory '/data/jiangyv/tp/kaldi/src/cudafeat' Makefile:127: recipe for target 'cudafeat/depend' failed make: [cudafeat/depend] Error 2 `
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/kaldi-asr/kaldi/issues/4252#issuecomment-684531002, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAZFLO5A4WJGNIHLFJPNVHDSDSSY7ANCNFSM4QRFNTTQ .
Actually @kkm this may be due to recent changes to the configure script. The relevant option is -I.. (or maybe -I ..)
On Tue, Sep 1, 2020 at 4:03 PM Daniel Povey dpovey@gmail.com wrote:
Probably because the current directory "." is not on the list of includes. Perhaps you did not run configure. Did you look at the INSTALL instructions?
On Tue, Sep 1, 2020 at 4:00 PM shanhaidexiamo notifications@github.com wrote:
Downloaded from where? How? … <#m_-8578085194426421633_m3399741452031424692> On Tue, Sep 1, 2020 at 2:57 PM jtrmal @.> wrote: maybe try "make clean" before make depend, just to make sure... Plus, we saw several issues with virtual machines using SMB volumes (window shares) y, On Tue, Sep 1, 2020 at 7:50 AM shanhaidexiamo @.> wrote: > I meet a similar problem when make depend and the error is: > > online-ivector-feature-cuda-kernels.cu:19:58: fatal error: > cudafeat/online-ivector-feature-cuda-kernels.h: No such file or directory >
include "cudafeat/online-ivector-feature-cuda-kernels.h" > > Besides,
there is no any error when I back to former version. > > I am sure this is a series errors, so your problem is same as me. Besides, > I find this problem in a new server, when I turn back to old server, > everything is ok
— > You are receiving this because you are subscribed to this thread. > Reply to this email directly, view it on GitHub > <#4252 (comment) https://github.com/kaldi-asr/kaldi/issues/4252#issuecomment-684375857>, or unsubscribe > < https://github.com/notifications/unsubscribe-auth/ACUKYXYJ6ZB437A5IDLHEE3SDSDTHANCNFSM4QRFNTTQ . > — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#4252 (comment) https://github.com/kaldi-asr/kaldi/issues/4252#issuecomment-684491543>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAZFLO4M3RMDG26YYAT22B3SDSLPLANCNFSM4QRFNTTQ .
just use git clone https://github.com/kaldi-asr/kaldi.git https://github.com/kaldi-asr/kaldi.git and then compile tools ,and then src. but when make depend in src, there have a problem : `#$ SPACE=
$ CUDART=cudart
$ HERE=/usr/local/cuda/bin
$ THERE=/usr/local/cuda/bin
$ TARGET_SIZE=
$ TARGET_DIR=
$ TARGET_SIZE=64
$ TOP=/usr/local/cuda/bin/..
$ NVVMIR_LIBRARY_DIR=/usr/local/cuda/bin/../nvvm/libdevice
$ LD_LIBRARY_PATH=/usr/local/cuda/bin/../lib:
$
PATH=/usr/local/cuda/bin/../nvvm/bin:/usr/local/cuda/bin:/data/jiangyv/anaconda3/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games
$ INCLUDES="-I/usr/local/cuda/bin/..//include"
$ LIBRARIES= "-L/usr/local/cuda/bin/..//lib64/stubs"
"-L/usr/local/cuda/bin/..//lib64"
$ CUDAFE_FLAGS=
$ PTXAS_FLAGS=
$ g++ -std=c++14 -D__CUDA_ARCH__=300 -E -x c++
-DCUDA_DOUBLE_MATH_FUNCTIONS -DCUDACC -DNVCC -I"/usr/local/cuda/include" -I"/data/jiangyv/tp/kaldi/tools/cub-1.8.0" "-I/usr/local/cuda/bin/..//include" -D "HAVE_CUDA" -D "KALDI_DOUBLEPRECISION=0" -D "CUDA_API_PER_THREAD_DEFAULT_STREAM" -DCUDACC_VER_MAJOR=10 -DCUDACC_VER_MINOR=0 -DCUDACC_VER_BUILD=130 -include "cuda_runtime.h" -m64 " online-ivector-feature-cuda-kernels.cu" > "/tmp/tmpxft_00003be6_00000000-4_online-ivector-feature-cuda-kernels.cpp1.ii" online-ivector-feature-cuda-kernels.cu:19:58: fatal error: cudafeat/online-ivector-feature-cuda-kernels.h: No such file or directory compilation terminated. --error 0x1 --
../makefiles/default_rules.mk:149: recipe for target 'depend' failed make[1]: [depend] Error 1 make[1]: Leaving directory '/data/jiangyv/tp/kaldi/src/cudafeat' Makefile:127: recipe for target 'cudafeat/depend' failed make: [cudafeat/depend] Error 2 `
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/kaldi-asr/kaldi/issues/4252#issuecomment-684531002, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAZFLO5A4WJGNIHLFJPNVHDSDSSY7ANCNFSM4QRFNTTQ .
Anyway, show your kaldi.mk, that will help track it down. There may have been changes in how flags are passed to the CUDA compiler.
On Tue, Sep 1, 2020 at 4:06 PM Daniel Povey dpovey@gmail.com wrote:
Actually @kkm this may be due to recent changes to the configure script. The relevant option is -I.. (or maybe -I ..)
On Tue, Sep 1, 2020 at 4:03 PM Daniel Povey dpovey@gmail.com wrote:
Probably because the current directory "." is not on the list of includes. Perhaps you did not run configure. Did you look at the INSTALL instructions?
On Tue, Sep 1, 2020 at 4:00 PM shanhaidexiamo notifications@github.com wrote:
Downloaded from where? How? … <#m_-2932421053838196097m-8578085194426421633_m3399741452031424692> On Tue, Sep 1, 2020 at 2:57 PM jtrmal @.> wrote: maybe try "make clean" before make depend, just to make sure... Plus, we saw several issues with virtual machines using SMB volumes (window shares) y, On Tue, Sep 1, 2020 at 7:50 AM shanhaidexiamo @.> wrote: > I meet a similar problem when make depend and the error is: > > online-ivector-feature-cuda-kernels.cu:19:58: fatal error: > cudafeat/online-ivector-feature-cuda-kernels.h: No such file or directory >
include "cudafeat/online-ivector-feature-cuda-kernels.h" > > Besides,
there is no any error when I back to former version. > > I am sure this is a series errors, so your problem is same as me. Besides, > I find this problem in a new server, when I turn back to old server, > everything is ok
— > You are receiving this because you are subscribed to this thread. > Reply to this email directly, view it on GitHub > <#4252 (comment) https://github.com/kaldi-asr/kaldi/issues/4252#issuecomment-684375857>, or unsubscribe > < https://github.com/notifications/unsubscribe-auth/ACUKYXYJ6ZB437A5IDLHEE3SDSDTHANCNFSM4QRFNTTQ . > — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#4252 (comment) https://github.com/kaldi-asr/kaldi/issues/4252#issuecomment-684491543>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAZFLO4M3RMDG26YYAT22B3SDSLPLANCNFSM4QRFNTTQ .
just use git clone https://github.com/kaldi-asr/kaldi.git https://github.com/kaldi-asr/kaldi.git and then compile tools ,and then src. but when make depend in src, there have a problem : `#$ SPACE=
$ CUDART=cudart
$ HERE=/usr/local/cuda/bin
$ THERE=/usr/local/cuda/bin
$ TARGET_SIZE=
$ TARGET_DIR=
$ TARGET_SIZE=64
$ TOP=/usr/local/cuda/bin/..
$ NVVMIR_LIBRARY_DIR=/usr/local/cuda/bin/../nvvm/libdevice
$ LD_LIBRARY_PATH=/usr/local/cuda/bin/../lib:
$
PATH=/usr/local/cuda/bin/../nvvm/bin:/usr/local/cuda/bin:/data/jiangyv/anaconda3/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games
$ INCLUDES="-I/usr/local/cuda/bin/..//include"
$ LIBRARIES= "-L/usr/local/cuda/bin/..//lib64/stubs"
"-L/usr/local/cuda/bin/..//lib64"
$ CUDAFE_FLAGS=
$ PTXAS_FLAGS=
$ g++ -std=c++14 -D__CUDA_ARCH__=300 -E -x c++
-DCUDA_DOUBLE_MATH_FUNCTIONS -DCUDACC -DNVCC -I"/usr/local/cuda/include" -I"/data/jiangyv/tp/kaldi/tools/cub-1.8.0" "-I/usr/local/cuda/bin/..//include" -D "HAVE_CUDA" -D "KALDI_DOUBLEPRECISION=0" -D "CUDA_API_PER_THREAD_DEFAULT_STREAM" -DCUDACC_VER_MAJOR=10 -DCUDACC_VER_MINOR=0 -DCUDACC_VER_BUILD=130 -include "cuda_runtime.h" -m64 " online-ivector-feature-cuda-kernels.cu" > "/tmp/tmpxft_00003be6_00000000-4_online-ivector-feature-cuda-kernels.cpp1.ii" online-ivector-feature-cuda-kernels.cu:19:58: fatal error: cudafeat/online-ivector-feature-cuda-kernels.h: No such file or directory compilation terminated. --error 0x1 --
../makefiles/default_rules.mk:149: recipe for target 'depend' failed make[1]: [depend] Error 1 make[1]: Leaving directory '/data/jiangyv/tp/kaldi/src/cudafeat' Makefile:127: recipe for target 'cudafeat/depend' failed make: [cudafeat/depend] Error 2 `
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/kaldi-asr/kaldi/issues/4252#issuecomment-684531002, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAZFLO5A4WJGNIHLFJPNVHDSDSSY7ANCNFSM4QRFNTTQ .
无论如何,请显示您的kaldi.mk,这将有助于对其进行跟踪。将标志传递到CUDA编译器的方式可能有所更改。 … 2020年9月1日星期二下午4:06丹尼尔·鲍维@.>写道:实际上@kkm这可能是由于最近对configure脚本所做的更改。相关选项是-I ..(或可能是-I ..),在2020年9月1日,星期二,下午4:03 Daniel Povey @.>写道:>可能是因为当前目录“。” 不在>包含列表中。也许您没有运行configure。您是否查看了INSTALL>说明?>>在2020年9月1日,星期二,下午4:00 shanhaidexiamo @.>>写道:> >>从哪里下载?怎么样?>>…>> <#m_-2932421053838196097m-8578085194426421633_m3399741452031424692> >>在2020年9月1日,星期二,jtrmal @.>写道:可能在makedepend之前尝试“ make >> clean”,只是为了确保...另外,我们在使用SMB卷(窗口共享)的虚拟机上看到了几个问题>>在2020年9月1日星期二>> 2020年7点:50 AM shanhaidexiamo @。>写道:>我遇到一个类似的问题>>当makedepend时,错误是:> >> >> online-ivector-feature-cuda-kernels.cu:19:58:致命错误:>> > cudafeat / online-ivector-feature-cuda-kernels.h:无此类文件或目录> >> #include“ cudafeat / online-ivector-feature-cuda-kernels.h”>>此外,>>没有任何内容回到旧版本时出现错误。>>我确定这是>>系列错误,因此您的问题与我相同。此外,>在新服务器中发现此>>问题,当我回到旧服务器时,>一切正常>>>>->您收到此消息是因为您已订阅此线程。> >>直接回复此电子邮件,在GitHub上查看> < 我遇到类似的问题>>当makedepend时,错误为:> >> >> online-ivector-feature-cuda-kernels.cu:19:58:致命错误:> >> cudafeat / online-ivector-feature-cuda -kernels.h:没有这样的文件或目录> >> #include“ cudafeat / online-ivector-feature-cuda-kernels.h”>>此外,>>回到我的旧版本时没有任何错误。>>我确定这是>>系列错误,因此您的问题与我相同。此外,>在新服务器中发现此>>问题,当我回到旧服务器时,>一切正常>>>>->您收到此消息是因为您已订阅此线程。> >>直接回复此电子邮件,在GitHub上查看> < 我遇到类似的问题>>当makedepend时,错误为:> >> >> online-ivector-feature-cuda-kernels.cu:19:58:致命错误:> >> cudafeat / online-ivector-feature-cuda -kernels.h:没有这样的文件或目录> >> #include“ cudafeat / online-ivector-feature-cuda-kernels.h”>>此外,>>回到我的旧版本时没有任何错误。>>我确定这是>>系列错误,因此您的问题与我相同。此外,>在新服务器中发现此>>问题,当我回到旧服务器时,>一切正常>>>>->您收到此消息是因为您已订阅此线程。> >>直接回复此电子邮件,在GitHub上查看> < online-ivector-feature-cuda-kernels.cu:19:58:致命错误:> >> cudafeat / online-ivector-feature-cuda-kernels.h:无此类文件或目录> >> #include“ cudafeat / online -ivector-feature-cuda-kernels.h“>>另外,当我回到以前的版本时,没有任何错误。>>我确定这是>>系列错误,因此您的问题与我相同。此外,>在新服务器中发现此>>问题,当我回到旧服务器时,>一切正常>>>>->您收到此消息是因为您已订阅此线程。> >>直接回复此电子邮件,在GitHub上查看> < online-ivector-feature-cuda-kernels.cu:19:58:致命错误:> >> cudafeat / online-ivector-feature-cuda-kernels.h:无此类文件或目录> >> #include“ cudafeat / online -ivector-feature-cuda-kernels.h“>>另外,当我回到以前的版本时,没有任何错误。>>我确定这是>>系列错误,因此您的问题与我相同。此外,>在新服务器中发现此>>问题,当我回到旧服务器时,>一切正常>>>>->您收到此消息是因为您已订阅此线程。> >>直接回复此电子邮件,在GitHub上查看> < 您收到此消息是因为您已订阅此线程。> >>直接回复此电子邮件,在GitHub上查看> < 您收到此消息是因为您已订阅此线程。> >>直接回复此电子邮件,在GitHub上查看> <#4252(评论)>> < #4252(评论) >>,>>>或取消订阅> <>> https://github.com/notifications/unsubscribe-auth/ACUKYXYJ6ZB437A5IDLHEE3SDSDTHANCNFSM4QRFNTTQ >>>>。> —您收到此评论是因为您发表了评论。直接回复此电子邮件>>,在GitHub < #4252(评论)>> < #4252(评论) >>,>>或取消订阅>> https://github.com/notifications/unsubscribe-auth/AAZFLO4M3RMDG26YYAT22B3SDSLPLANCNFSM4QRFNTTQ >>。>> >>只需使用 git clone https://github.com/kaldi-asr/kaldi.git >> < https://github.com/kaldi-asr/kaldi.git > [](https://github.com/kaldi-asr/kaldi.git>)然后编译工具,然后>>然后是src。但是当src中的 makedepend 时,有一个问题:>>`#$ SPACE = >>#$ CUDART = cudart >>#$ HERE = / usr / local / cuda / bin >> #$ 那里 = / usr / local / cuda / bin >>#$ TARGET_SIZE = >>#$ TARGET_DIR = >>#$ TARGET_SIZE = 64 >>#$ TOP = / usr / local / cuda / bin / .. >>#$ NVVMIR_LIBRARY_DIR = / usr / local / cuda / bin /../ nvvm / libdevice >>#$ LD_LIBRARY_PATH = / usr / local / cuda / bin /../ lib:>># $ >> PATH = / usr / local / cuda / bin /../ nvvm / bin:/ usr / local / cuda / bin:/ data / jiangyv / anaconda3 / bin:/ usr / local / sbin:/ usr / local / bin:/ usr / sbin:/ usr / bin:/ sbin:/ bin:/ usr / games:/ usr / local / games >>#$ INCLUDES =“-I / usr / local / cuda / bin / .. // include“ >>#$ LIBRARIES =” -L / usr / local / cuda / bin /..// lib64 / stubs“ >>” -L / usr / local / cuda / bin / .. cudafeat / online-ivector-feature-cuda-kernels.h:没有此类文件或目录>>终止。>> --error 0x1->> >> ../makefiles/default_rules.mk:149:目标“依赖”的配方失败>> make [1]:* [depend]错误1 >> make [1 ]:离开目录'/ data / jiangyv / tp / kaldi / src / cudafeat'>> Makefile:127:目标'cudafeat / depend'的配方失败>> make: [cudafeat / depend]错误2 >>` >> >> — >>您收到此评论是因为您发表了评论。>>直接回复此电子邮件,在GitHub上查看>> [depend]错误1 >> make [1]:离开目录'/ data / jiangyv / tp / kaldi / src / cudafeat'>> Makefile:127:目标'cudafeat / depend'的配方失败>> make: [cudafeat / depend]错误2 >>`>> >> — >>您收到此消息是因为您评论了。>>直接回复此电子邮件,在GitHub上查看>> [depend]错误1 >> make [1]:离开目录'/ data / jiangyv / tp / kaldi / src / cudafeat'>> Makefile:127:目标'cudafeat / depend'的配方失败>> make: *** [cudafeat / depend]错误2 >>`>> >> — >>您收到此消息是因为您评论了。>>直接回复此电子邮件,在GitHub上查看>>#4252(注释) >,>>或取消订阅>> < https://github.com/notifications/unsubscribe-auth/AAZFLO5A4WJGNIHLFJPNVHDSDSSY7ANCNFSM4QRFNTTQ > >>。>>>
it seems that MKL part is different from the former version , is that the reason?
same error for me today.
same error for me today.
did you find the method?
no, actually I am cloning kaldi with a specific commit, but I hope that it will be fixed with this .
no, actually I am cloning kaldi with a specific commit, but I hope that it will be fixed with this .
I just use the kaldi-pyblind11 version to avoid this problem. I find that the historical version has no problem.
Fixed now by merging that commit. The -I.. was definitely the issue. I think Hugo had removed the standard CXXFLAGS from the CUDA invocation somehow.
On Thu, Sep 3, 2020 at 1:10 AM shanhaidexiamo notifications@github.com wrote:
no, actually I am cloning kaldi with a specific commit, but I hope that it will be fixed with this https://github.com/kaldi-asr/kaldi/pull/4250 .
I just use the kaldi-pyblind11 version to avoid this problem. I find that the historical version has no problem.
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/kaldi-asr/kaldi/issues/4252#issuecomment-685874287, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAZFLO2IOI74C6GCEBJMQPLSDZ37XANCNFSM4QRFNTTQ .
Fixed now by merging that commit. The -I.. was definitely the issue. I think Hugo had removed the standard CXXFLAGS from the CUDA invocation somehow. … On Thu, Sep 3, 2020 at 1:10 AM shanhaidexiamo @.***> wrote: no, actually I am cloning kaldi with a specific commit, but I hope that it will be fixed with this <#4250> . I just use the kaldi-pyblind11 version to avoid this problem. I find that the historical version has no problem. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#4252 (comment)>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAZFLO2IOI74C6GCEBJMQPLSDZ37XANCNFSM4QRFNTTQ .
Thanks for your reply! make depend is going well. Next step is doing make, and I find the Linkbase problem. It seems that a series problems. Maybe the configure file still have sth wrong or there's something wrong with my server environment. Hope no one else has had this problem
Can you show us the exact args you used with configure
and the contents
of kaldi.mk?
On Thu, Sep 3, 2020 at 1:55 PM shanhaidexiamo notifications@github.com wrote:
Fixed now by merging that commit. The -I.. was definitely the issue. I think Hugo had removed the standard CXXFLAGS from the CUDA invocation somehow. … <#m-8625416860088786311> On Thu, Sep 3, 2020 at 1:10 AM shanhaidexiamo @.***> wrote: no, actually I am cloning kaldi with a specific commit, but I hope that it will be fixed with this <#4250 https://github.com/kaldi-asr/kaldi/pull/4250> . I just use the kaldi-pyblind11 version to avoid this problem. I find that the historical version has no problem. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#4252 (comment) https://github.com/kaldi-asr/kaldi/issues/4252#issuecomment-685874287>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAZFLO2IOI74C6GCEBJMQPLSDZ37XANCNFSM4QRFNTTQ .
Thanks for your reply! make depend is going well. Next step is doing make, and I find the Linkbase problem. It seems that a series problems. Maybe the configure file still have sth wrong or there's something wrong with my server environment. Hope no one else has had this problem [image: 微信截图_20200903135515] https://user-images.githubusercontent.com/48786488/92076431-1eeb7180-eded-11ea-9c85-4aa1f9233677.png
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/kaldi-asr/kaldi/issues/4252#issuecomment-686271344, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAZFLO3EP6IODLSZKQPLS5TSD4VWXANCNFSM4QRFNTTQ .
for shared -fPIC should definitely be in the CXXFLAGS...
On Thu, Sep 3, 2020 at 8:09 AM shanhaidexiamo notifications@github.com wrote:
Can you show us the exact args you used with configure and the contents of kaldi.mk? On Thu, Sep 3, 2020 at 1:55 PM shanhaidexiamo notifications@github.com wrote: … <#m2865677302196190472> Fixed now by merging that commit. The -I.. was definitely the issue. I think Hugo had removed the standard CXXFLAGS from the CUDA invocation somehow. … <#m-8625416860088786311> On Thu, Sep 3, 2020 at 1:10 AM shanhaidexiamo @.**> wrote: no, actually I am cloning kaldi with a specific commit, but I hope that it will be fixed with this <#4250 https://github.com/kaldi-asr/kaldi/pull/4250 <#4250 https://github.com/kaldi-asr/kaldi/pull/4250>> . I just use the kaldi-pyblind11 version to avoid this problem. I find that the historical version has no problem. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#4252 https://github.com/kaldi-asr/kaldi/issues/4252 (comment) <#4252 (comment) https://github.com/kaldi-asr/kaldi/issues/4252#issuecomment-685874287>>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAZFLO2IOI74C6GCEBJMQPLSDZ37XANCNFSM4QRFNTTQ . Thanks for your reply! make depend is going well. Next step is doing make*, and I find the Linkbase problem. It seems that a series problems. Maybe the configure file still have sth wrong or there's something wrong with my server environment. Hope no one else has had this problem [image: 微信截图_20200903135515] https://user-images.githubusercontent.com/48786488/92076431-1eeb7180-eded-11ea-9c85-4aa1f9233677.png — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#4252 (comment) https://github.com/kaldi-asr/kaldi/issues/4252#issuecomment-686271344>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAZFLO3EP6IODLSZKQPLS5TSD4VWXANCNFSM4QRFNTTQ .
just ./configure --shared. and here is my kaldi.mk and I did not change the configure file. I just run some commands.
`# This file was generated using the following command: ./configure --shared
CONFIGURE_VERSION := 11 Toolchain configuration
CXX = g++ AR = ar AS = as RANLIB = ranlib Base configuration
KALDI_FLAVOR := dynamic KALDILIBDIR := /data/jiangyv/old/kaldi-master/src/lib DEBUG_LEVEL = 1 DOUBLE_PRECISION = 0 OPENFSTINC = /data/jiangyv/old/kaldi-master/tools/openfst-1.6.7/include OPENFSTLIBS = /data/jiangyv/old/kaldi-master/tools/openfst-1.6.7/lib/libfst.so OPENFSTLDFLAGS = -Wl,-rpath=/data/jiangyv/old/kaldi-master/tools/openfst-1.6.7/lib
CUBROOT = /data/jiangyv/old/kaldi-master/tools/cub-1.8.0 WITH_CUDADECODER = 1
MKLROOT = /opt/intel/mkl MKLLIB = /opt/intel/mkl/lib/intel64 MKL specific Linux configuration We have tested Kaldi with MKL version 10.2 on Linux/GCC and Intel(R) 64 architecture (also referred to as x86_64) with LP64 interface layer. The linking flags for MKL will be very different depending on the OS, architecture, compiler, etc. used. The correct flags can be obtained from http://software.intel.com/en-us/articles/intel-mkl-link-line-advisor/ Use the options obtained from this website to manually configure for other platforms using MKL.
ifndef DEBUG_LEVEL $(error DEBUG_LEVEL not defined.) endif ifndef DOUBLE_PRECISION $(error DOUBLE_PRECISION not defined.) endif ifndef OPENFSTINC $(error OPENFSTINC not defined.) endif ifndef OPENFSTLIBS $(error OPENFSTLIBS not defined.) endif ifndef MKLROOT $(error MKLROOT not defined.) endif
MKLLIB ?= $(MKLROOT)/lib/intel64
CXXFLAGS = -std=c++11 -I.. -isystem $(OPENFSTINC) -O1 $(EXTRA_CXXFLAGS) -Wall -Wno-sign-compare -Wno-unused-local-typedefs -Wno-deprecated-declarations -Winit-self -DKALDI_DOUBLEPRECISION=$(DOUBLE_PRECISION) -DHAVE_EXECINFO_H=1 -DHAVE_CXXABI_H -DHAVE_MKL -I$(MKLROOT)/include -m64 -msse -msse2 -pthread -g
ifeq ($(KALDI_FLAVOR), dynamic) CXXFLAGS += -fPIC endif
ifeq ($(DEBUG_LEVEL), 0) CXXFLAGS += -DNDEBUG endif ifeq ($(DEBUG_LEVEL), 2) CXXFLAGS += -O0 -DKALDI_PARANOID endif Compiler specific flags
COMPILER = $(shell $(CXX) -v 2>&1) ifeq ($(findstring clang,$(COMPILER)),clang) Suppress annoying clang warnings that are perfectly valid per spec.
CXXFLAGS += -Wno-mismatched-tags endif
LDFLAGS = $(EXTRA_LDFLAGS) $(OPENFSTLDFLAGS) -rdynamic LDLIBS = $(EXTRA_LDLIBS) $(OPENFSTLIBS) $(MKLFLAGS) -lm -lpthread -ldl MKLFLAGS = -L/opt/intel/mkl/lib/intel64 -Wl,-rpath=/opt/intel/mkl/lib/intel64 -lmkl_intel_lp64 -lmkl_core -lmkl_sequential -ldl -lpthread -lm CUDA configuration
CUDA = true CUDATKDIR = /usr/local/cuda CUDA_ARCH = -gencode arch=compute_30,code=sm_30 -gencode arch=compute_35,code=sm_35 -gencode arch=compute_50,code=sm_50 -gencode arch=compute_52,code=sm_52 -gencode arch=compute_60,code=sm_60 -gencode arch=compute_61,code=sm_61 -gencode arch=compute_70,code=sm_70 -gencode arch=compute_75,code=sm_75
ifndef DOUBLE_PRECISION $(error DOUBLE_PRECISION not defined.) endif ifndef CUDATKDIR $(error CUDATKDIR not defined.) endif
CXXFLAGS += -DHAVE_CUDA -I$(CUDATKDIR)/include -fPIC -pthread -isystem $(OPENFSTINC)
CUDA_INCLUDE= -I$(CUDATKDIR)/include -I$(CUBROOT) -I.. -isystem $(OPENFSTINC) CUDA_FLAGS = --machine 64 -DHAVE_CUDA -ccbin $(CXX) -DKALDI_DOUBLEPRECISION=$(DOUBLE_PRECISION) -std=c++14 -DCUDA_API_PER_THREAD_DEFAULT_STREAM -lineinfo --verbose -Wno-deprecated-gpu-targets
ifeq ($(shell test -e $(CUDATKDIR)/lib64/libcudart_static.a && echo -n yes),yes) CUDA_LDFLAGS += -L$(CUDATKDIR)/lib64 -Wl,-rpath,$(CUDATKDIR)/lib64 else CUDA_LDFLAGS += -L$(CUDATKDIR)/lib -Wl,-rpath,$(CUDATKDIR)/lib endif
CUDA_LDLIBS += -lcublas -lcusparse -lcudart -lcurand -lcufft -lnvToolsExt
LDLIBS : The .so libs are loaded later than static libs in implicit rule
CUDA_LDLIBS += -lcusolver Environment configuration
`
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/kaldi-asr/kaldi/issues/4252#issuecomment-686276477, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACUKYX4SBZDSLPB5VCGAOUDSD4XKFANCNFSM4QRFNTTQ .
It is, it needs to be in CUDA_FLAGS though. We removed CXXFLAGS from the cuda flags. I think makefiles/cuda_64bit.mk need to be edited, then rerun configure. PR please!
On Thu, Sep 3, 2020 at 2:12 PM jtrmal notifications@github.com wrote:
for shared -fPIC should definitely be in the CXXFLAGS...
On Thu, Sep 3, 2020 at 8:09 AM shanhaidexiamo notifications@github.com wrote:
Can you show us the exact args you used with configure and the contents of kaldi.mk? On Thu, Sep 3, 2020 at 1:55 PM shanhaidexiamo notifications@github.com wrote: … <#m2865677302196190472> Fixed now by merging that commit. The -I.. was definitely the issue. I think Hugo had removed the standard CXXFLAGS from the CUDA invocation somehow. … <#m-8625416860088786311> On Thu, Sep 3, 2020 at 1:10 AM shanhaidexiamo @.***> wrote: no, actually I am cloning kaldi with a specific commit, but I hope that it will be fixed with this <#4250 https://github.com/kaldi-asr/kaldi/pull/4250 <#4250 https://github.com/kaldi-asr/kaldi/pull/4250>> . I just use the kaldi-pyblind11 version to avoid this problem. I find that the historical version has no problem. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#4252 https://github.com/kaldi-asr/kaldi/issues/4252 (comment) <#4252 (comment) <https://github.com/kaldi-asr/kaldi/issues/4252#issuecomment-685874287
, or unsubscribe
https://github.com/notifications/unsubscribe-auth/AAZFLO2IOI74C6GCEBJMQPLSDZ37XANCNFSM4QRFNTTQ . Thanks for your reply! make depend is going well. Next step is doing make, and I find the Linkbase problem. It seems that a series problems. Maybe the configure file still have sth wrong or there's something wrong with my server environment. Hope no one else has had this problem [image: 微信截图_20200903135515]
https://user-images.githubusercontent.com/48786488/92076431-1eeb7180-eded-11ea-9c85-4aa1f9233677.png — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#4252 (comment) <https://github.com/kaldi-asr/kaldi/issues/4252#issuecomment-686271344
, or unsubscribe
https://github.com/notifications/unsubscribe-auth/AAZFLO3EP6IODLSZKQPLS5TSD4VWXANCNFSM4QRFNTTQ .
just ./configure --shared. and here is my kaldi.mk and I did not change the configure file. I just run some commands.
`# This file was generated using the following command: ./configure --shared
CONFIGURE_VERSION := 11 Toolchain configuration
CXX = g++ AR = ar AS = as RANLIB = ranlib Base configuration
KALDI_FLAVOR := dynamic KALDILIBDIR := /data/jiangyv/old/kaldi-master/src/lib DEBUG_LEVEL = 1 DOUBLE_PRECISION = 0 OPENFSTINC = /data/jiangyv/old/kaldi-master/tools/openfst-1.6.7/include OPENFSTLIBS = /data/jiangyv/old/kaldi-master/tools/openfst-1.6.7/lib/libfst.so OPENFSTLDFLAGS = -Wl,-rpath=/data/jiangyv/old/kaldi-master/tools/openfst-1.6.7/lib
CUBROOT = /data/jiangyv/old/kaldi-master/tools/cub-1.8.0 WITH_CUDADECODER = 1
MKLROOT = /opt/intel/mkl MKLLIB = /opt/intel/mkl/lib/intel64 MKL specific Linux configuration We have tested Kaldi with MKL version 10.2 on Linux/GCC and Intel(R) 64 architecture (also referred to as x86_64) with LP64 interface layer. The linking flags for MKL will be very different depending on the OS, architecture, compiler, etc. used. The correct flags can be obtained from http://software.intel.com/en-us/articles/intel-mkl-link-line-advisor/ Use the options obtained from this website to manually configure for other platforms using MKL.
ifndef DEBUG_LEVEL $(error DEBUG_LEVEL not defined.) endif ifndef DOUBLE_PRECISION $(error DOUBLE_PRECISION not defined.) endif ifndef OPENFSTINC $(error OPENFSTINC not defined.) endif ifndef OPENFSTLIBS $(error OPENFSTLIBS not defined.) endif ifndef MKLROOT $(error MKLROOT not defined.) endif
MKLLIB ?= $(MKLROOT)/lib/intel64
CXXFLAGS = -std=c++11 -I.. -isystem $(OPENFSTINC) -O1 $(EXTRA_CXXFLAGS) -Wall -Wno-sign-compare -Wno-unused-local-typedefs -Wno-deprecated-declarations -Winit-self -DKALDI_DOUBLEPRECISION=$(DOUBLE_PRECISION) -DHAVE_EXECINFO_H=1 -DHAVE_CXXABI_H -DHAVE_MKL -I$(MKLROOT)/include -m64 -msse -msse2 -pthread -g
ifeq ($(KALDI_FLAVOR), dynamic) CXXFLAGS += -fPIC endif
ifeq ($(DEBUG_LEVEL), 0) CXXFLAGS += -DNDEBUG endif ifeq ($(DEBUG_LEVEL), 2) CXXFLAGS += -O0 -DKALDI_PARANOID endif Compiler specific flags
COMPILER = $(shell $(CXX) -v 2>&1) ifeq ($(findstring clang,$(COMPILER)),clang) Suppress annoying clang warnings that are perfectly valid per spec.
CXXFLAGS += -Wno-mismatched-tags endif
LDFLAGS = $(EXTRA_LDFLAGS) $(OPENFSTLDFLAGS) -rdynamic LDLIBS = $(EXTRA_LDLIBS) $(OPENFSTLIBS) $(MKLFLAGS) -lm -lpthread -ldl MKLFLAGS = -L/opt/intel/mkl/lib/intel64 -Wl,-rpath=/opt/intel/mkl/lib/intel64 -lmkl_intel_lp64 -lmkl_core -lmkl_sequential -ldl -lpthread -lm CUDA configuration
CUDA = true CUDATKDIR = /usr/local/cuda CUDA_ARCH = -gencode arch=compute_30,code=sm_30 -gencode arch=compute_35,code=sm_35 -gencode arch=compute_50,code=sm_50 -gencode arch=compute_52,code=sm_52 -gencode arch=compute_60,code=sm_60 -gencode arch=compute_61,code=sm_61 -gencode arch=compute_70,code=sm_70 -gencode arch=compute_75,code=sm_75
ifndef DOUBLE_PRECISION $(error DOUBLE_PRECISION not defined.) endif ifndef CUDATKDIR $(error CUDATKDIR not defined.) endif
CXXFLAGS += -DHAVE_CUDA -I$(CUDATKDIR)/include -fPIC -pthread -isystem $(OPENFSTINC)
CUDA_INCLUDE= -I$(CUDATKDIR)/include -I$(CUBROOT) -I.. -isystem $(OPENFSTINC) CUDA_FLAGS = --machine 64 -DHAVE_CUDA -ccbin $(CXX) -DKALDI_DOUBLEPRECISION=$(DOUBLE_PRECISION) -std=c++14 -DCUDA_API_PER_THREAD_DEFAULT_STREAM -lineinfo --verbose -Wno-deprecated-gpu-targets
ifeq ($(shell test -e $(CUDATKDIR)/lib64/libcudart_static.a && echo -n yes),yes) CUDA_LDFLAGS += -L$(CUDATKDIR)/lib64 -Wl,-rpath,$(CUDATKDIR)/lib64 else CUDA_LDFLAGS += -L$(CUDATKDIR)/lib -Wl,-rpath,$(CUDATKDIR)/lib endif
CUDA_LDLIBS += -lcublas -lcusparse -lcudart -lcurand -lcufft -lnvToolsExt
LDLIBS : The .so libs are loaded later than static libs in implicit rule
CUDA_LDLIBS += -lcusolver Environment configuration
`
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/kaldi-asr/kaldi/issues/4252#issuecomment-686276477, or unsubscribe < https://github.com/notifications/unsubscribe-auth/ACUKYX4SBZDSLPB5VCGAOUDSD4XKFANCNFSM4QRFNTTQ
.
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/kaldi-asr/kaldi/issues/4252#issuecomment-686277515, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAZFLO4HTMADNFZSDDFYBLLSD4XUHANCNFSM4QRFNTTQ .
Can you show us the exact args you used with
configure
and the contents of kaldi.mk? On Thu, Sep 3, 2020 at 1:55 PM shanhaidexiamo notifications@github.com wrote: … Fixed now by merging that commit. The -I.. was definitely the issue. I think Hugo had removed the standard CXXFLAGS from the CUDA invocation somehow. … <#m-8625416860088786311> On Thu, Sep 3, 2020 at 1:10 AM shanhaidexiamo @.**> wrote: no, actually I am cloning kaldi with a specific commit, but I hope that it will be fixed with this <#4250 <#4250>> . I just use the kaldi-pyblind11 version to avoid this problem. I find that the historical version has no problem. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#4252 (comment) <#4252 (comment)>>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAZFLO2IOI74C6GCEBJMQPLSDZ37XANCNFSM4QRFNTTQ . Thanks for your reply! make depend is going well. Next step is doing make*, and I find the Linkbase problem. It seems that a series problems. Maybe the configure file still have sth wrong or there's something wrong with my server environment. Hope no one else has had this problem [image: 微信截图_20200903135515] https://user-images.githubusercontent.com/48786488/92076431-1eeb7180-eded-11ea-9c85-4aa1f9233677.png — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#4252 (comment)>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAZFLO3EP6IODLSZKQPLS5TSD4VWXANCNFSM4QRFNTTQ .
I just use the command "./configure --shared" and I did not change any configure files. The code is briefly shown here, I removed some annotations:
CONFIGURE_VERSION := 11
CXX = g++ AR = ar AS = as RANLIB = ranlib
KALDI_FLAVOR := dynamic KALDILIBDIR := /data/jiangyv/old/kaldi-master/src/lib DEBUG_LEVEL = 1 DOUBLE_PRECISION = 0 OPENFSTINC = /data/jiangyv/old/kaldi-master/tools/openfst-1.6.7/include OPENFSTLIBS = /data/jiangyv/old/kaldi-master/tools/openfst-1.6.7/lib/libfst.so OPENFSTLDFLAGS = -Wl,-rpath=/data/jiangyv/old/kaldi-master/tools/openfst-1.6.7/lib
CUBROOT = /data/jiangyv/old/kaldi-master/tools/cub-1.8.0 WITH_CUDADECODER = 1
MKLROOT = /opt/intel/mkl MKLLIB = /opt/intel/mkl/lib/intel64
ifndef DEBUG_LEVEL $(error DEBUG_LEVEL not defined.) endif ifndef DOUBLE_PRECISION $(error DOUBLE_PRECISION not defined.) endif ifndef OPENFSTINC $(error OPENFSTINC not defined.) endif ifndef OPENFSTLIBS $(error OPENFSTLIBS not defined.) endif ifndef MKLROOT $(error MKLROOT not defined.) endif
MKLLIB ?= $(MKLROOT)/lib/intel64
CXXFLAGS = -std=c++11 -I.. -isystem $(OPENFSTINC) -O1 $(EXTRA_CXXFLAGS) \ -Wall -Wno-sign-compare -Wno-unused-local-typedefs \ -Wno-deprecated-declarations -Winit-self \ -DKALDI_DOUBLEPRECISION=$(DOUBLE_PRECISION) \ -DHAVE_EXECINFO_H=1 -DHAVE_CXXABI_H -DHAVE_MKL -I$(MKLROOT)/include \ -m64 -msse -msse2 -pthread \ -g
ifeq ($(KALDI_FLAVOR), dynamic) CXXFLAGS += -fPIC endif
ifeq ($(DEBUG_LEVEL), 0) CXXFLAGS += -DNDEBUG endif ifeq ($(DEBUG_LEVEL), 2) CXXFLAGS += -O0 -DKALDI_PARANOID endif
COMPILER = $(shell $(CXX) -v 2>&1) ifeq ($(findstring clang,$(COMPILER)),clang)
CXXFLAGS += -Wno-mismatched-tags endif
LDFLAGS = $(EXTRA_LDFLAGS) $(OPENFSTLDFLAGS) -rdynamic LDLIBS = $(EXTRA_LDLIBS) $(OPENFSTLIBS) $(MKLFLAGS) -lm -lpthread -ldl MKLFLAGS = -L/opt/intel/mkl/lib/intel64 -Wl,-rpath=/opt/intel/mkl/lib/intel64 -lmkl_intel_lp64 -lmkl_core -lmkl_sequential -ldl -lpthread -lm
CUDA = true CUDATKDIR = /usr/local/cuda CUDA_ARCH = -gencode arch=compute_30,code=sm_30 -gencode arch=compute_35,code=sm_35 -gencode arch=compute_50,code=sm_50 -gencode arch=compute_52,code=sm_52 -gencode arch=compute_60,code=sm_60 -gencode arch=compute_61,code=sm_61 -gencode arch=compute_70,code=sm_70 -gencode arch=compute_75,code=sm_75
ifndef DOUBLE_PRECISION $(error DOUBLE_PRECISION not defined.) endif ifndef CUDATKDIR $(error CUDATKDIR not defined.) endif
CXXFLAGS += -DHAVE_CUDA -I$(CUDATKDIR)/include -fPIC -pthread -isystem $(OPENFSTINC)
CUDA_INCLUDE= -I$(CUDATKDIR)/include -I$(CUBROOT) -I.. -isystem $(OPENFSTINC) CUDA_FLAGS = --machine 64 -DHAVE_CUDA \ -ccbin $(CXX) -DKALDI_DOUBLEPRECISION=$(DOUBLE_PRECISION) \ -std=c++14 -DCUDA_API_PER_THREAD_DEFAULT_STREAM -lineinfo \ --verbose -Wno-deprecated-gpu-targets
ifeq ($(shell test -e $(CUDATKDIR)/lib64/libcudart_static.a && echo -n yes),yes) CUDA_LDFLAGS += -L$(CUDATKDIR)/lib64 -Wl,-rpath,$(CUDATKDIR)/lib64 else CUDA_LDFLAGS += -L$(CUDATKDIR)/lib -Wl,-rpath,$(CUDATKDIR)/lib endif
CUDA_LDLIBS += -lcublas -lcusparse -lcudart -lcurand -lcufft -lnvToolsExt #LDLIBS : The .so libs are loaded later than static libs in implicit rule CUDA_LDLIBS += -lcusolver
Got the same error as @shanhaidexiamo.
Got the same error as @shanhaidexiamo.
you mean the Linkbase problem or the make depend problem?
Linkbase
Linkbase
all right, maybe the configure file or makefiles still need to be updated.
As I said, you need to add the -fPIC flags to makefiles/cuda_64bit.mk, reconfigure, and try again. Please make a PR if you have time.
On Thu, Sep 3, 2020 at 4:11 PM shanhaidexiamo notifications@github.com wrote:
Linkbase [image: Schermata 2020-09-03 alle 10 07 36] https://user-images.githubusercontent.com/7140210/92088656-6e21aa00-edcd-11ea-9bf5-189fa839768a.png
all right, maybe the configure file or makefiles still need to be updated.
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/kaldi-asr/kaldi/issues/4252#issuecomment-686329823, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAZFLO4R7YKIZ4IMIZ26XKLSD5FSBANCNFSM4QRFNTTQ .
Add --compiler-options -fPIC
to CUDA_FLAGS
in makefiles/cuda_64bit.mk may solve the problem.
Now it's working, thank you.
online-ivector-feature-cuda-kernels.cu:19:10: fatal error: cudafeat/online-ivector-feature-cuda-kernels.h: No such file or directory
#include "cudafeat/online-ivector-feature-cuda-kernels.h"
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
compilation terminated.
I am getting this error when running make depend
. I'm using --shared
and passing a cudatk-dir. Latest kaldi (am using -fPIC for CUDA). My CUDA version is 10.2
This issue has been automatically marked as stale by a bot solely because it has not had recent activity. Please add any comment (simply 'ping' is enough) to prevent the issue from being closed for 60 more days if you believe it should be kept open.
make depend is leading to shit loads of errors inside src folder.
Whosoever is forcing others to make changes on their own is an idiot.
Thank you for your opinion. We send a keychain as a token of appreciation. y.
On Thu, Jan 27, 2022 at 10:18 AM dexception @.***> wrote:
make depend is leading to shit loads of errors inside src folder.
Whosoever is forcing others to make changes on their own is an idiot.
— Reply to this email directly, view it on GitHub https://github.com/kaldi-asr/kaldi/issues/4252#issuecomment-1023001852, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACUKYX7JVCJJGZMGTHHY6ELUYEEUZANCNFSM4QRFNTTQ . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.
You are receiving this because you commented.Message ID: @.***>
This issue has been automatically marked as stale by a bot solely because it has not had recent activity. Please add any comment (simply 'ping' is enough) to prevent the issue from being closed for 60 more days if you believe it should be kept open.
When I try to make depend in src, there always sth wrong , such as "feature-online-cmvn-cuda.cu:19:47: fatal error: cudafeat/feature-online-cmvn-cuda.h: No such file or directory". And I find that these files are internal files in kaldi, they failed when they include each other. On the other hand, it does not report errors when I change include "cudafeat/feature-online-cmvn-cuda.h" to an absolute path, but that's not a good idea. Can someone tell me what the problem is and how to solve it? Thank you !