git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit

]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit

overview / pkg / ggml / sources / llama.cpp / commit

author	Georgi Gerganov <redacted>
	Fri, 23 Jan 2026 16:22:34 +0000 (18:22 +0200)
committer	GitHub <redacted>
	Fri, 23 Jan 2026 16:22:34 +0000 (18:22 +0200)
commit	557515be1e93ed8939dd8a7c7d08765fdbe8be31
tree	85ea47a16ae0097d197fc472bed2f2e653601896	tree
parent	cb6caca191b9a3a9a4eaa13dd9e465225d127034	commit \| diff

graph : utilize `ggml_build_forward_select()` to avoid reallocations (#18898)

* graph : avoid branches between embedding and token inputs

* models : make deepstack graphs (e.g. Qwen3 VL) have constant topology

* ci : enable -DGGML_SCHED_NO_REALLOC=ON for server CI

* cont : pad token embeddings to n_embd_inp

Packaging of ggml-org/llama.cpp

RSS Atom

.github/workflows/server.yml		diff \| blob \| history
src/llama-context.cpp		diff \| blob \| history
src/llama-graph.cpp		diff \| blob \| history
src/llama-graph.h		diff \| blob \| history
src/models/gemma3n-iswa.cpp		diff \| blob \| history
src/models/qwen3vl-moe.cpp		diff \| blob \| history
src/models/qwen3vl.cpp		diff \| blob \| history