5 ms·
HIP is only useful to convert CUDA source code. Tensorflow uses the cuDNN library, which is closed source. There is nothing for HIP to convert.
by TomVDB 9y ago
HIP is only useful to convert CUDA source code.
Tensorflow uses the cuDNN library, which is closed source. There is nothing for HIP to convert.
- tgtweak 9y agoAMD is working on a cudnn comptability layer iirc (MiOpen) and is ROCm group has created a cuda transpiler (to intermediary HIP then to amd binary via hcc) https://instinct.radeon.com/en/6-deep-learning-projects-amd-radeon-instinct/ https://instinct.radeon.com/en/6-deep-learning-projects-amd-...
- TomV1971 9y agoYes, but it means AMD will have to do the original work. They can’t use a CUDA to HIP converter.
- mastazi 9y agoWait, I thought that AMD's cuDNN replacement was hipDNN, now I'm confused https://github.com/ROCmSoftwarePlatform/hipDNN https://github.com/ROCmSoftwarePlatform/hipDNN Edit: OK I was reading the docs, I think I got it: hipDNN is a wrapper that (once finished) will search and replace calls (from cuDNN to hipDNN), then hipDNN itself, in turn, will call MIOpen, not sure if that's right, I would appreciate if someone who knows more could confirm
- nl 9y agoAMD is always working on random half-assed attempts to get to parity with NVidia for neural network training. If they'd chosen one approach 5 years ago and put decent resources behind it they might be competitive by now.