-
Note: Providing complete information in the most concise form is the best way to get help. This issue template serves as the checklist for essential information to most of the technical issues and bug reports. For non-technical issues and feature requests, feel free to present the information in what you believe is the best form. For Q & A and discussion, please start a discussion thread at https://discuss.mxnet.io Description(Brief description of the problem in no more than 2 sentences.) Environment info (Required)Platform: Power9 We tried to install the mxnet-cu101 package for python2.7 on our powerpc machine, but pip cannot find the version:
Thank you! |
Beta Was this translation helpful? Give feedback.
Replies: 15 comments
-
Hey, this is the MXNet Label Bot. |
Beta Was this translation helpful? Give feedback.
-
Sorry, but we are not officially supporting PowerPC. Please try compiling yourself. |
Beta Was this translation helpful? Give feedback.
-
@mxnet-label-bot add [question] |
Beta Was this translation helpful? Give feedback.
-
@marcoabreu Hello, is there any guide we can follow for building mxnet on power9 ? |
Beta Was this translation helpful? Give feedback.
-
You may follow https://mxnet.apache.org/get_started/ubuntu_setup in analogy |
Beta Was this translation helpful? Give feedback.
-
Yeah, this guide helped us too.
|
Beta Was this translation helpful? Give feedback.
-
If the guide needs any extension to cover PowerPC, feel free to open a PR. Thanks |
Beta Was this translation helpful? Give feedback.
-
@leezu @Talgin , Thanks for helping guys, It worked ! But there is little problem i got, my progress takes 1 sec on 1080ti on 64 bit intel, 1.2 sec on v100 on ppc64. v100 on ppc should be faster than 1080ti on 64 bit. You guys have any idea ? |
Beta Was this translation helpful? Give feedback.
-
You need to investigate and find some minimal example that shows the problem. For example, by using the benchmark scripts https://github.com/apache/incubator-mxnet/tree/master/benchmark/opperf |
Beta Was this translation helpful? Give feedback.
-
@pasa13142, we had the same problem, training was slower on power9 with Tesla V100 than on my workstation. I do not remember if they found a reason, but our server management service said that ppc was not originally made to do such things (training) :) Like they are good at calculation, maybe some kind of server operations, but they (ibms ppc) do not have much experience with GPUs. |
Beta Was this translation helpful? Give feedback.
-
Thank you, if i find out, i'll update, but unfortunately signs are not good on ibms ppc :( |
Beta Was this translation helpful? Give feedback.
-
@Talgin hey! Which math library did you select for powerpc ? And does it work or does it matter for gpu ? what could be make v100 slower than 1080 ? :/ |
Beta Was this translation helpful? Give feedback.
-
Hi, @pasa13142! We used numpy, scikit-learn, pandas and many other. Installation was without problems and they worked OK. Have you found something interesting? :) |
Beta Was this translation helpful? Give feedback.
-
Hi, @pasa13142 |
Beta Was this translation helpful? Give feedback.
You may follow https://mxnet.apache.org/get_started/ubuntu_setup in analogy