| Index | index by Group | index by Distribution | index by Vendor | index by creation date | index by Name | Mirrors | Help |
The search service can find package by either name (apache), provides(webserver), absolute file names (/usr/bin/apache), binaries (gprof) or shared libraries (libXm.so.2) in standard path. It does not support multiple arguments yet...
The System and Arch are optional added filters, for example System could be "redhat", "redhat-7.2", "mandrake" or "gnome", Arch could be "i386" or "src", etc. depending on your system.
torchvulkan registers a Vulkan device with PyTorch and runs tensor ops as Vulkan compute shaders. Importing it claims the "vulkan" device name (PrivateUse1). That name then takes priority over the in-tree ATen Vulkan backend built into python-torch. Most operators run on the GPU. The ones that do not fall back to the CPU, including during training. TORCHVULKAN_STRICT=1 turns those fallbacks into errors. Shaders are written in Slang, compiled to SPIR-V when this package is built, and embedded in the extension. A Vulkan 1.3 driver is needed at runtime. volk is bundled and loads the system Vulkan loader; VulkanMemoryAllocator is the bundled header.
| Package | Summary | Distribution | Download |
| python-torchvulkan-0.2.1-1.aarch64.html | Vulkan compute backend for PyTorch | OpenMandriva Cooker for aarch64 | python-torchvulkan-0.2.1-1.aarch64.rpm |
| python-torchvulkan-0.2.1-1.x86_64.html | Vulkan compute backend for PyTorch | OpenMandriva Cooker for x86_64 | python-torchvulkan-0.2.1-1.x86_64.rpm |
Generated by rpm2html 1.6