Implementation of algorithms for memory optimized deep neural network training
-
Updated
Jul 23, 2020 - Cuda
Implementation of algorithms for memory optimized deep neural network training
Budget-aware Vulkan memory allocation for V, with block suballocation, upload rings, fragmentation diagnostics, and sustained-load validation.
To associate your repository with the memory-budget topic, visit your repo's landing page and select "manage topics."