Upgrade llama.cpp from b9022 to b9049#104
Merged
bernardladenthin merged 1 commit intomasterfrom May 7, 2026
Merged
Conversation
Key changes in this range: - New LLAMA_STATE_SEQ_FLAGS_ON_DEVICE flag for on-device KV cache save/restore - State seq data format now prepends 4-byte magic + seq_id header (b9022 state data incompatible) - ggml_op_hint enum + ggml_mul_mat_set_hint() for FWHT support - llama_backend_init() auto-loads backends if none registered - server_prompt_checkpoint_update() gained on_device parameter - GGML version 0.10.2 → 0.11.0, cpp-httplib 0.43.2 → 0.43.3 https://claude.ai/code/session_01TZ2Gvm2dyeRVoy5dCzMXNn
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
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
This PR upgrades the pinned llama.cpp dependency from version b9022 to b9049, incorporating upstream improvements and new features for KV cache state management, FWHT support, and backend initialization.
Key Changes
KV Cache State Management: New
LLAMA_STATE_SEQ_FLAGS_ON_DEVICEflag enables on-device KV cache state save/restore without host round-trips. State data format now includes 4-byte magic header and seq_id, making saved state from b9022 incompatible with b9049+.FWHT Support: New
ggml_op_hintenum andggml_mul_mat_set_hint()function added for Fast Walsh-Hadamard Transform support in graph operations.Backend Initialization:
llama_backend_init()now automatically callsggml_backend_load_all()if no backends are registered, simplifying initialization flow.Error Handling: Unsupported model architectures now throw
std::runtime_errorinstead of callingGGML_ABORT, allowing graceful error handling by callers.Speculative Decoding: Server context checkpoints now support on-device state flags for improved speculative decoding performance.
Dependencies: GGML version bumped from 0.10.2 to 0.11.0; cpp-httplib updated to 0.43.3 with recursion elimination and OOM-safe improvements.
Notes
No JNI layer call-site changes are required. The new on-device state features and FWHT hints are not currently used by the Java bindings but are available for future optimization.
https://claude.ai/code/session_01TZ2Gvm2dyeRVoy5dCzMXNn