You are not logged in.

#1 2022-05-08 21:38:33

feng
Member
Registered: 2019-03-26
Posts: 10

tensorflow-opt-cuda: can it support legacy GPU K80 with sm_37?

I was testing a small tensorflow/keras application with my old K80 card, but I noticed that this package is built up starting from sm_52.
Can the maintainer enable sm_37 sm_50 if possible? I assume only need to update the line `export TF_CUDA_COMPUTE_CAPABILITIES=sm_52,sm_53,sm_60,sm_61,sm_62,sm_70,sm_72,sm_75,sm_80,sm_86,compute_86` to `export TF_CUDA_COMPUTE_CAPABILITIES=sm_37,sm_50,sm_52,sm_53,sm_60,sm_61,sm_62,sm_70,sm_72,sm_75,sm_80,sm_86,compute_86`, and there will be no significant compile time increasing.

Offline

#2 2022-05-08 21:48:04

Slithery
Forum Moderator
From: Norfolk, UK
Registered: 2013-12-01
Posts: 5,241

Re: tensorflow-opt-cuda: can it support legacy GPU K80 with sm_37?

The developers rarely visit the forums. Posting a feature request on the bug tracker against the package is a better way of getting their attention.


No, it didn't "fix" anything. It just shifted the brokeness one space to the right. - jasonwryan
Closing -- for deletion; Banning -- for muppetry. - jasonwryan

aur - dotfiles

Online

#3 2022-05-09 05:26:39

feng
Member
Registered: 2019-03-26
Posts: 10

Re: tensorflow-opt-cuda: can it support legacy GPU K80 with sm_37?

Slithery wrote:

The developers rarely visit the forums. Posting a feature request on the bug tracker against the package is a better way of getting their attention.

Many thanks for reminding me of this. I will do it.

Offline

Board footer

Powered by FluxBB