Our presentation from the GPU Technology Conference is now available for streaming; we continue testing AIR-T prototypes; delivery has been pushed back to February; and TensorRT 5.0 and Windows 10 support has been added to GR-Wavelearner.
Back in late October, Deepwave gave an invited talk to the NVIDIA GPU Technology Conference - one of the premier AI conferences centered around high-performance computing. The 50-minute presentation and corresponding white paper are now available:
The Deepwave team has been testing AIR-T prototypes and we are making great progress. Over the previous months, we have eliminated the possibility of major issues and we are now optimizing the interfaces between the FPGA, GPU, and RF front end. We were expecting to leverage the vendor supplied reference designs for some of the drivers, but these have proven to require a significant amount of enhancements to work with the AIR-T. For that reason, we must push out the delivery of the AIR-T until February 28, 2019. We hope to be completed sooner and will keep you posted on our progress.
Up until now, GR-Wavelearner supported TensorRT 3.3, the version compatible with the NVIDIA Tegra TX2. By request, we've added Windows 10 support and, with that, we brought in the latest TensorRT libraries. Give it a try and let us know what you think. GR-Wavelearner can be found here.
Thanks for your support,
The Deepwave Team