feat: update llama.cpp to a956a852c by abetlen · Pull Request #2414 · abetlen/llama-cpp-python

abetlen · 2026-11-14T18:15:16Z

Update the vendored llama.cpp from 7d7f02875 to a956a852c and synchronize the Python bindings.

Bind the public speculative-decoding APIs and updated MTMD video/audio interfaces.
Preserve deprecated Python entry points and existing MTP extension imports using current native APIs.
Validation: Linux CPU build, all 26 tests, Ruff, native/ctypes ABI comparison, and compatibility/MTMD resampling smoke tests passed.
GPU speculative decoding and end-to-end custom video decoding were not tested locally.

abetlen added 2 commits November 14, 2026 10:15

feat: update llama.cpp to a956a852c

05a143b

docs: record llama.cpp update and align MTMD comments

28a7f17

abetlen marked this pull request as ready for review November 14, 2026 18:18

abetlen added 2 commits November 14, 2026 10:52

docs: match DFlash enum comments to upstream header

7a3fd30

docs: consolidate unreleased llama.cpp update entry

2ddf018

abetlen merged commit dd70c92 into main Nov 14, 2026
15 checks passed

Navigation Menu