Bump llama.cpp to 8452824 (b9739), release v0.8.27#59
Merged
Conversation
Update vendor/llama.cpp from 74ade5274 to 845282461 (67 commits, tag b9739). No NIF changes required: all binding-relevant headers are unchanged except common/speculative.h (two new optional get/set_state declarations the NIF does not call) and common/common.h (changes to common_params / common_params_model, which the NIF never constructs — it builds only common_params_speculative, whose sole change is internal need_n_rs_seq() logic now also covering EAGLE3 drafts). Verified against a freshly built NIF: 158 tests + 4 skipped, all 7 end-to-end smoke tests pass (generation, streaming, chat templates, JSON-schema grammar, raw GBNF, and embeddings), mix format clean, Dialyzer 0 errors.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Bumps the
vendor/llama.cppsubmodule from74ade5274to845282461(67 commits, tag b9739) and cuts release v0.8.27.
No NIF changes were required. Of the headers the binding compiles against:
include/llama.h,ggml/include/ggml.h,ggml/include/ggml-backend.h,common/chat.h,common/json-schema-to-grammar.h, andcommon/sampling.hare unchanged.
common/speculative.honly gains two optional declarations —common_speculative_get_state/common_speculative_set_state(stash/restoreinternal speculative state) — which the binding does not call.
swaps its
namefield for aget_name()method, the deprecatedwebui/webui_mcp_proxy/webui_config_jsonfields are dropped fromcommon_params, amodels_preset_hffield and anfs_open_ifstreamhelperare added, and
common_prompt_checkpointgains adata_specblob. The NIFconstructs only
common_params_speculative(settingtypesanddraft.*),never
common_paramsorcommon_params_model, and the solecommon_params_speculativechange is internalneed_n_rs_seq()logic thatnow also reserves a recurrent-state seq for EAGLE3 drafts.
Notable upstream changes
long-prompt segfault fix (#24707).
rope_parametershandling (#24833); skipmain_gpuvalidation when nodevices are available (#23405).
operator (#24724),
rope_backoperator (#24725).#24778), preprocessor refactor (#24736).
#24760, #24739, #23976), invalid-grammar HTTP 400 (#24154).
updates. See
CHANGELOG.mdfor the full list.Verification
Against a freshly rebuilt NIF (Metal, Apple M4 Max):
mix test— 158 passed, 4 skippedJSON-schema grammar, raw GBNF, and embeddings — embedding paths fully
exercised with a Qwen3-Embedding-0.6B model)
mix format --check-formatted— cleanmix dialyzer— 0 errorsNotes
checksum.exsis intentionally not updated — CI regenerates it againstthe precompiled release artifacts after the
v0.8.27tag is pushed.v0.8.27onmasterto trigger the precompile + checksumworkflow.