]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
llama: avoid copying logits during prompt decode in MTP (#23198)
authorAman Gupta <redacted>
Sun, 17 May 2026 15:30:25 +0000 (23:30 +0800)
committerGitHub <redacted>
Sun, 17 May 2026 15:30:25 +0000 (23:30 +0800)
commit3e12fbdea5c1ac4225c7dcf79506d30950283fc3
treefdfcf376e4e5973327812d79b51e29b326f9e7d4
parent39cf5d61915769124b7efbbfa69c46f19a6363ee
llama: avoid copying logits during prompt decode in MTP (#23198)

* llama: avoid copying logits during prompt decode in MTP

* review: update comment

* llama-graph: call set_output for t_h_pre_norm
common/speculative.cpp
common/speculative.h
src/llama-context.cpp
src/llama-context.h
src/llama-cparams.h
src/llama-ext.h
src/llama-graph.cpp
src/models/qwen35.cpp
src/models/qwen35moe.cpp
tools/server/server-context.cpp