OpenSource/ollama

Auteur	SHA1 Message	Date
Jeffrey Morgan	6164f378f2 revert cuda overhead to 20%	il y a 1 an
Jeffrey Morgan	6566387ae3 add `TODO` for cuda overhead	il y a 1 an
Jeffrey Morgan	37708931fb update cuda overhead to 20% to fix crashes when switching between models and large context sizes	il y a 1 an
Jeffrey Morgan	f6cb0a553c update cuda overhead to 15% or 400MiB	il y a 1 an
Jeffrey Morgan	2680078c13 fix build on linux	il y a 1 an
Jeffrey Morgan	f1b7e5f560 update overhead to 15%	il y a 1 an
Jeffrey Morgan	cb534e6ac2 use 10% vram overhead for cuda	il y a 1 an
Jeffrey Morgan	08f1e18965 Offload layers to GPU based on new model size estimates (#1850)	il y a 1 an
Daniel Hiltgen	d74ce6bd4f Detect very old CUDA GPUs and fall back to CPU	il y a 1 an
Daniel Hiltgen	a2ad952440 Fix windows system memory lookup	il y a 1 an
Daniel Hiltgen	d966b730ac Switch windows build to fully dynamic	il y a 1 an
Daniel Hiltgen	7555ea44f8 Revamp the dynamic library shim	il y a 1 an
Daniel Hiltgen	1b991d0ba9 Refine build to support CPU only	il y a 1 an
Daniel Hiltgen	35934b2e05 Adapted rocm support to cgo based llama.cpp	il y a 1 an