llama.cpp
llama : disable graph reuse with pipeline parallelism
#20463

Merged

llama : disable graph reuse with pipeline parallelism #20463

ggerganov merged 2 commits into master from gg/llama-disable-graph-reuse-with-pp

llama : disable graph reuse with pipeline parallelism

dfa3ad18

ggerganov force pushed from 7bc73aef to dfa3ad18 59 days ago

Revert "CUDA: Improve performance via less synchronizations between …

14be4a46

github-actions added Nvidia GPU

github-actions added ggml

JohannesGaessler approved these changes on 2026-03-12

ORippler approved these changes on 2026-03-12

ggerganov merged 57819b8d into master 59 days ago

ggerganov deleted the gg/llama-disable-graph-reuse-with-pp branch 59 days ago

Reviewers

ORippler

JohannesGaessler

Assignees

No one assigned

Labels

Nvidia GPU ggml

Milestone

No milestone