Draft: ggml-opencl: Early proof-of-concept implementation of plans via command buffers #22764
hacky patches to make it work on pocl-cuda
138d41ec
Implement and use cuda graph plans.
9abbfe0a
opencl: use command buffers when available
03423817
opencl: async-ify tensor I/O a bit
6049b971
hacks to make the ROCm compiler happy(ier)
5b85b9bf
Assignees
No one assigned
Login to write a write a comment.
Login via GitHub