The Vulkan spec does not enforce strict IEEE 754 floating point semantics, so perfect compatibility with CUDA is impossible. https://registry.khronos.org/vulka

The Vulkan spec does not enforce strict IEEE 754 floating point semantics, so perfect compatibility with CUDA is impossible. https://registry.khronos.org/vulkan/specs/1.3-khr-extensions/html/chap43.html#spirvenv-op-prec https://registry.khronos.org/vulkan/specs/1.3-khr-extensions... However, the deep learning field does currently not pay much attention to reproducibility, so this might not be a big issue.