Deep Learning on ROCm

I wonder if it will have a positive impact on the development of this library that the new Mac Pro will feature a dedicated AMD Vega GPU.<p>EDIT: there is more discussion going on at &#x2F;r&#x2F;MachineLearning than here: <a href="https:&#x2F;&#x2F;www.reddit.com&#x2F;r&#x2F;MachineLearning&#x2F;comments&#x2F;6kv3rs&#x2F;mopen_10_released_by_amd_deep_learning_software&#x2F;" rel="nofollow">https:&#x2F;&#x2F;www.reddit.com&#x2F;r&#x2F;MachineLearning&#x2F;comments&#x2F;6kv3rs&#x2F;mop...</a>

Are there any benchmarks comparing performance with CuDNN?

&gt; the first open-source HPC&#x2F;Hyperscale-class platform for GPU computing that’s also programming-language independent<p>Weird website, took me a while to classify its purpose. So, is this like an AMD version of NVIDIA&#x27;s CuDNN? Can we run TensorFlow on AMD GPUs?