pip install error: No matching distribution found for mxnet-cu80 #6023
Comments
What's the output of this command?
|
@szha |
@ysh329 There should be versions that are available then. Could it be that there are more than one copy of pip that's on your system, and that the one invoked by sudo is different from the 9.0.1 one? Just to make sure, could you try to invoke the installation command using absolute path of the pip that returned version 9.0.1? |
@szha I only installed one pip. Besides, I switched to root and pip 9.0.1 from /usr/lib/python2.7/site-packages (python2.7) |
0.9.5 pip packages have been released for a while and should have resolved this problem. |
@ysh329 has this issue been resolved? Also, we have released 0.10.0 so this is the version you will get from pip now. |
@szha Ok, I'll try. Besides, I tried to use |
@ysh329 has the issue been fixed for you? |
@szha Due to poor network, I don't know 😞 |
@ysh329 Sorry to hear that... |
I'm in exact same issue with this. My environment is Nvidia TX1 board, and they equipped cu80. |
@wangmir are you using centos 7 too? Unfortunately, I don't have a box to test it on centos... |
nop, I'm using ubuntu 16.04 on TX1. |
What's your pip version? |
my pip version is 9.0.1 and I also did pip install --upgrade pip |
@wangmir what's the output of |
this :: |
@wangmir I see. Thanks for the info. The problem is that I only have x86_64 built so far. Let me look into how to get aarch64. |
Ok, Thanks. At first, I'll try to build from source. |
I tried to build the source on my own hand, but it fails too. Issue #6846. |
I am having the same issues with my machine |
@TalKachman yes, you can build from source, and then go to python folder, and use the anaconda's python to do a |
@szha Ubuntu 16.04 and python >3 |
@TalKachman you should be able to use the binary release since we have a manylinux version of it for mxnet-cu80. Did you upgrade your pip to > 8.1? |
@szha
I ended up compiling it from source with the GPU support, and compilation works well but when importing money it is still broken
|
I am having the same issues with my machineCould not find a version that satisfies the requirement setup.py (from versions: ) please help me on this.. |
@punisher-n can't help without knowing what your system is |
For those who still have the installation issue, please use the script here provided by @zhreshold to report the environment information for better diagnosis, and open new issues. Thanks. |
Install this one |
I am getting the same installation error for kera on windows 10. pip version is |
I got the similar error and my system is ppc64le. |
Same error on macOS 10.13:
With Python 2.7 I can install version 1.2.1post1, but it doesn't upgrade to 1.3.0. On Ubuntu 18 it works fine. |
@mariussoutier I'd like to see what pep425 tags are supported on your mac. What's the output of import pip
pip.pep425tags.get_supported() Note that w/ pip10/18 the above may fail, and you can instead do import pip._internal
pip._internal.pep425tags.get_supported() |
Both fail, actually. I have pip 18. |
weird. mine seems to work: % ipython
Python 3.7.0 (v3.7.0:1bf9cc5093, Jun 26 2018, 23:26:24)
Type 'copyright', 'credits' or 'license' for more information
IPython 6.4.0 -- An enhanced Interactive Python. Type '?' for help.
In [1]: import pip
In [2]: pip.__version__
Out[2]: '18.0'
In [3]: import pip._internal
In [4]: pip._internal.pep425tags.get_supported()
Out[4]:
[('cp37', 'cp37m', 'macosx_10_13_x86_64'),
('cp37', 'cp37m', 'macosx_10_13_intel'), Do you mind running this script and reporting the output? This is for collecting some environment information. |
Ah sorry, the latter works, I didn't import
|
@mariussoutier looks like the problem is that we didn't release the 3.7 support yet. For lower versions your python only supports ('cp36', 'none') whereas we have (cp36, cp36m, macosx_10_11_x86_64). I will see if I can quickly release one (allow 2 hours). However, note that 3.7 support hasn't been incorporated in mxnet's CI yet. If you run into issue, let us know immediately so that we can at least quickly fix in the master branch (and nightly). |
Ok, |
@szha Can we have pip build from source when it encounters unsupported python / arch, e.g., arm? |
@yzhliu that's possible if we do a source distribution. however, we'd have to assume that the build config we use is what the user wants. |
I install mxnet-cu80 on CentOS7 (Power Linux) , which has been installed cuda.
The text was updated successfully, but these errors were encountered: