`ggml_cuda_cpy` support for 4d tensors and float16->float32 upcasting#686

Merged

slaren merged 7 commits intoggml-org:masterfrom balisujohn:dev-cuda-4d-copy

Jan 29, 2024

+89-50

Commits on Jan 7, 2024

added cuda float16->float32 upcasting to ggml_cuda_cpy
balisujohn
committed

Commits on Jan 8, 2024

added ability to copy 4d tensors with the cuda backend
balisujohn
committed

Commits on Jan 12, 2024

Commits on Jan 29, 2024