General purpose GPU compute framework built on Vulkan to support 1000s of cross vendor graphics cards (AMD, Qualcomm, NVIDIA & friends). Blazing fast, mobile-enabled, asynchronous and optimized for advanced GPU data processing usecases. Backed by the Linux Foundation.
Currently it's not possible to specify the region to copy from when calling the tensor.recordCopyFrom function. Should investigate the most effective way in which this can be exposed in case people may want to extend this functionality.
Bump, I am interested in this.
It would be nice if it was possible to sync only a part of a tensor with the device/host.
Also, specifying tensor offsets for a shader dispatch, is that possible in Vulkan?
Currently it's not possible to specify the region to copy from when calling the tensor.recordCopyFrom function. Should investigate the most effective way in which this can be exposed in case people may want to extend this functionality.