Package Details: caffe-cuda 1.0-12

Git Clone URL: (read-only, click to copy)
Package Base: caffe-cuda
Description: A deep learning framework made with expression, speed, and modularity in mind (with cuda support)
Upstream URL:
Licenses: BSD
Conflicts: caffe
Provides: caffe
Submitter: dbermond
Maintainer: dbermond
Last Packager: dbermond
Votes: 2
Popularity: 0.000413
First Submitted: 2018-08-22 18:02 (UTC)
Last Updated: 2022-03-01 14:46 (UTC)

Latest Comments

milesxu commented on 2020-07-28 01:16 (UTC)

@dbermond I see, thank you! I found that CUDA has been updated to 11 in ArchLinux. But I am using Manjaro, so I will update caffe-cuda after Manjaro release CUDA 11 package.

dbermond commented on 2020-07-27 16:46 (UTC)

@milesxu It looks like that you have not updated your system (to get a new cuda version) before building the package. What's your cuda version?

milesxu commented on 2020-07-27 08:57 (UTC)

When build the newest version 1.0-10, the error below is occured: nvcc fatal : Unsupported gpu architecture 'compute_80'

ruffy commented on 2019-06-22 13:06 (UTC) (edited on 2019-06-22 13:07 (UTC) by ruffy)

@dbermond okay thanks

dbermond commented on 2019-06-22 12:01 (UTC) (edited on 2019-06-22 12:01 (UTC) by dbermond)

@ruffy The cuda package is showing issues when compiling cuda code after the gcc 8 -> 9 update. A ticket is opened at the bug tracker and this is being worked on by the cuda package maintainer.

gcc8 will be soon on the repos to fix this.

ruffy commented on 2019-06-22 07:09 (UTC) (edited on 2019-06-22 07:12 (UTC) by ruffy)

Anybody knows how to solve this issue?



usr/include/c++/9.1.0/bits/stl_function.h(437): error: identifier "__builtin_is_constant_evaluated" is undefined

1 error detected in the compilation of "/tmp/tmpxft_00006cb5_00000000-18_adam_solver.compute_75.cpp1.ii". make: *** [Makefile:608: .build_release/cuda/src/caffe/solvers/adam_solver.o] Error 1 ==> ERROR: A failure occurred in build(). Aborting...

dbermond commented on 2019-05-19 13:28 (UTC)

@ruffy Nice. Good to hear.

I think that error was due to some recent changes in texlive package. Now everything is fine.

ruffy commented on 2019-05-19 09:19 (UTC)

@dbermond, thx. I was able to install it, after quite some time.

I think that there is also some dependency missing. The error that the LaTeX package 'newunicodechar.sty' is missing occured during building.

I manually installed it on my machine, after that everything was fine.

dbermond commented on 2019-05-18 20:17 (UTC)

@ruffy Probably it's because it was still depending on gcc7. I've updated the package. Try now and it should be faster.

ruffy commented on 2019-05-10 06:44 (UTC) (edited on 2019-05-10 06:45 (UTC) by ruffy)

Is it normal, that the installation process needs more than 1.5 hour?

I never experienced such a long installation for any package in the AUR or official package.

I started yesterday the installation and after 1.5 hour I had to stop the installation, because I had no time to wait longer.

Is there anybody with the same experience?

dbermond commented on 2018-09-29 01:26 (UTC)

@s3v3n I do not use anaconda. Package is building fine.

s3v3n commented on 2018-09-27 19:03 (UTC) (edited on 2018-09-27 19:10 (UTC) by s3v3n)

I'm trying to solve an error at the last step. All dependencies are ok.

==> Building and installing package ==> Making package: caffe-cuda 1.0-1 (Thu 27 Sep 2018 03:50:49 PM -03) ==> Checking runtime dependencies... ==> Checking buildtime dependencies... ==> Retrieving sources... -> Found caffe-cuda-1.0.tar.gz -> Found Makefile.config ==> Validating source files with sha256sums... caffe-cuda-1.0.tar.gz ... Passed Makefile.config ... Passed ==> Extracting sources... -> Extracting caffe-cuda-1.0.tar.gz with bsdtar ==> Starting prepare()... Python 3.6.6 :: Anaconda custom (64-bit) ==> ERROR: A failure occurred in prepare(). Aborting... ==> ERROR: Makepkg was unable to build caffe-cuda. ==> Restart building caffe-cuda ? [y/N] ==> ----------------------------------

Is there a way to solve or workaround it?