This repository was archived by the owner on Aug 23, 2026. It is now read-only.
Repository navigation
Bump vendored llama.cpp b9585 -> b10068 - #68
Merged
Merged
Conversation
Re-vendor upstream release b10068 (2026-07-18). Trims match the
prior policy: only cpu/blas/metal/cuda ggml backends kept; no
docs/examples/tools/tests/scripts/conversion; new ggml-et backend
also trimmed.
Carries forward the two local patches on the same anchors:
- common/chat.{h,cpp}: skip_parser_synthesis flag on
common_chat_templates_inputs + render-only short-circuit in
common_chat_templates_apply_jinja. Backs the ParamsRef cache in
barrel_inference:chat_apply/2.
- include/llama.h + src/llama-model.{h,cpp}: prefetch field on
llama_model_params (POSIX_MADV_WILLNEED vs POSIX_MADV_RANDOM);
llama_model_n_mappings / llama_model_get_mapping C API exposing
the mmap regions. Backs weight_residency = lazy_then_pin_resident
and the resident_bytes gauge.
New upstream src/ additions are picked up automatically (models/
CMakeLists globs *.cpp; ggml-cuda globs *.cu/*.cuh). NIF builds
clean.
benoitc
added a commit
to benoitc/erllama
that referenced
this pull request
Aug 23, 2026
…ence#68) Re-vendor upstream release b10068 (2026-07-18). Trims match the prior policy: only cpu/blas/metal/cuda ggml backends kept; no docs/examples/tools/tests/scripts/conversion; new ggml-et backend also trimmed. Carries forward the two local patches on the same anchors: - common/chat.{h,cpp}: skip_parser_synthesis flag on common_chat_templates_inputs + render-only short-circuit in common_chat_templates_apply_jinja. Backs the ParamsRef cache in barrel_inference:chat_apply/2. - include/llama.h + src/llama-model.{h,cpp}: prefetch field on llama_model_params (POSIX_MADV_WILLNEED vs POSIX_MADV_RANDOM); llama_model_n_mappings / llama_model_get_mapping C API exposing the mmap regions. Backs weight_residency = lazy_then_pin_resident and the resident_bytes gauge. New upstream src/ additions are picked up automatically (models/ CMakeLists globs *.cpp; ggml-cuda globs *.cu/*.cuh). NIF builds clean.
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 subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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
common/chat.{h,cpp}:skip_parser_synthesisflag oncommon_chat_templates_inputs+ render-only short-circuit incommon_chat_templates_apply_jinja. Backs the ParamsRef cache inbarrel_inference:chat_apply/2.include/llama.h+src/llama-model.{h,cpp}:prefetchfield onllama_model_params;llama_model_n_mappings/llama_model_get_mappingC API exposing the mmap regions. Backsweight_residency = lazy_then_pin_residentand the resident_bytes gauge.models/CMakeLists globs*.cpp;ggml-cudaglobs*.cu/*.cuh). NIF builds clean.