github llama.cpp · just now · inference

llama.cpp b10662 adds ctx-per-slot for unified KV cache

from English

The llama.cpp project released build b10662, introducing a new `--kv-unified-per-slot` argument that allows users to specify context per slot for the unified KV cache. This change replaces the previous context fraction mechanism with context pool slots and refactors the underlying implementation.

Adds the `ctx-per-slot` argument to manage unified KV cache allocation more granularly.
Swaps out ctx fractions for ctx pool slots to improve memory management logic.
Removes the deprecated `ctx-pool-slots` parameter and converts `ctx-per-slot` to an integer type.
Includes general formatting cleanup and code refactoring by Xuan Son Nguyen.

This update provides users with finer control over context window distribution across multiple slots in the unified KV cache system.

Importance 1/3 Trust 1/3 llama.cpp Inference efficiency Local inference

Read original

github llama.cpp · 1h ago · 1 view Live

llama.cpp adds Qwen3.8-Flash-Next (qwen4exp) architecture support

The llama.cpp project has added support for the Qwen3.8-Flash-Next (qwen4exp) model architecture, implementing its specific components including hyper-connections, gated delta nets, Mixture of Experts, and PLE n-gram hash embeddings.

github llama.cpp · 2h ago · 1 view

llama.cpp b10658 adds DFlash2 support and fixes mrope bug

The llama.cpp project released build b10658, introducing support for the DFlash2 specification which includes local convolution and a candidate selector. This update also addresses a specific mrope bug within the codebase.

github llama.cpp · 2h ago · 1 view

llama.cpp bundles HIP runtime DLLs with Windows ROCm release

The llama.cpp project has updated its Windows ROCm release to bundle the necessary HIP runtime DLLs directly alongside the binaries. This change addresses issue #26929 by ensuring the correct HIP runtime loads instead of potentially conflicting with the driver's copy in System32.

github llama.cpp · 3h ago · 1 view

llama.cpp b10656 caps working memory to avoid loading big tensors onto RAM

The llama.cpp project has released build b10656, which includes a change to cap the working memory size. This modification is designed to prevent large tensors from being loaded into system RAM.

github llama.cpp · 3h ago · 1 view

llama.cpp b10657 adds OpenCL bin kernels for GEMO MOE

The llama.cpp project released version b10657, introducing new OpenCL binary kernels for mixed-expert models. Specifically, the update adds `kernel_gemm_moe_q4_0_q8_1_dp4a_bin` and `kernel_gemm_moe_mxfp4_q8_1_dp4a_bin` to support efficient computation on compatible hardware.