vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2025-09-10 17:14:36 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

switched to NTK aware scaling

Browse source
This commit is contained in:
Concedo 2023-07-02 17:25:08 +08:00
parent e19483ca0f
commit e17c8497cf
4 changed files with 26 additions and 25 deletions
Download patch file Download diff file Expand all files Collapse all files

2
llama.cpp
View file

@ -2633,7 +2633,7 @@ struct llama_context * llama_new_context_with_model(
ctx->buf_compute.resize(MEM_REQ_EVAL().at(ctx->model.type));
const size_t bigctxmul = (hparams.n_ctx>2048?2:1);
const size_t bigctxmul = (hparams.n_ctx>4096?3:(hparams.n_ctx>2048?2:1));
ctx->buf_scratch[0].resize(MEM_REQ_SCRATCH0().at(ctx->model.type)*bigctxmul);
ctx->buf_scratch[1].resize(MEM_REQ_SCRATCH1().at(ctx->model.type)*bigctxmul);
}