HSA

[visheshchanana]

I just bought a new Desktop with Ryzen 5 CPU and an AMD GPU to learn GPU programming. I am also interested in learning Tensorflow for deep neural networks. 

After a few days of fiddling with tensorflow on CPU, I realized I should shift all the computations to GPU. The tensorflow-gpu library isn't built for AMD as it uses CUDA while the openCL library cannot be used with tensorflow(I guess). 

Is there a library that I could use for tensorflow computations on AMD GPU??

[dipak]

As I know, AMD provides a ROCm enabled TensorFlow library for AMD GPUs. It's based on the ROCm software stack. To know more about this library, please find the below links:

https://github.com/ROCmSoftwarePlatform/tensorflow-upstream

https://gpuopen.com/rocm-tensorflow-1-8-release/

AMD also provides its own open source deep learning library, called MIOpen, for high performance machine learning primitives. Here are few useful links that give you more information about MIOpen library as well as ROCm software stack.

MIOpen:

https://gpuopen.com/compute-product/miopen/

https://github.com/ROCmSoftwarePlatform/MIOpen

ROCm:

https://rocm.github.io/documentation.html

GitHub - RadeonOpenCompute/ROCm: ROCm - Open Source Platform for HPC and Ultrascale GPU Computing 

https://github.com/ROCmSoftwarePlatform

For any support related to these libraries or ROCm software stack, please use corresponding GitHub sites. For example,  please use this GitHub link for any support related to TensorFlow library: https://github.com/ROCmSoftwarePlatform/tensorflow-upstream/issues

P.S. You have been whitelisted.

Thanks.