Re:AMD is determined to gets its rightful datacenter share
by lostmsu on Friday March 06, 2020 @03:17PM
(#22506910)
TensorFlow works with AMD's CUDA alternative named ROCm. You'd need to check installation instructions though.
[ Parent
| on HN ]
Re:AMD is determined to gets its rightful datacenter share
by jamesblonde on Friday March 06, 2020 @04:59PM
(#22507733)
Yes, ROCm is upstreamed into Tensorflow. On our Hopsworks platform, you can run the same TensorFlow code on either ROCm or Cuda. No code changes needed:
https://www.logicalclocks.com/blog/welcoming-amd-rocm-to-hop...
Performance ain't so bad on AMD GPUs anymore - and none of the EULA crap. Just buy a Vega R7 for 500 bucks and get 16 GB of memory and the performance of close to a 2080 Ti (for convnets, at least) - but with no data center restrictions:
https://github.com/ROCmSoftwarePlatform/tensorflow-upstream/...
Re:AMD is determined to gets its rightful datacenter share
by yangikan on Friday March 06, 2020 @08:21PM
(#22509034)
How about Pytorch?
Re:AMD is determined to gets its rightful datacenter share
by jamesblonde on Saturday March 07, 2020 @12:31AM
(#22509962)
You have to compile it yourself, which sucks. For some reason PyTorch aren't agreeing to upstream ROCm, yet.
Re:AMD is determined to gets its rightful datacenter share
by bgorman on Saturday March 07, 2020 @02:56AM
(#22510456)
You don't need to compile it, AMD provides docker images.
For example I was able to run fast.ai/pytorch oh my amd gpu https://github.com/briangorman/fastai_rocm_docker