gemma4: graph optimizations (massive gains on pp for E2B and E4B) by max-krasnyansky · Pull Request #31356 · ggml-org/llama.cpp
retained text
big prompt processing speedup for gemma E2B/E4B
document fetched 2026-11-04 00:41 · canonical https://reddit.com/r/LocalLLaMA/comments/1xn6mm0/gemma4_graph_optimizations_massive_gains_on_pp