github llama.cpp · just now · inference

llama.cpp b10435 fixes quadratic cost in jinja gather_string_parts

from English

The llama.cpp project has released version b10435, which addresses a performance issue within the Jinja template engine. The primary change fixes a quadratic time complexity bug in the `gather_string_parts` function.

Fixes quadratic cost in `gather_string_parts`
Updates comments and removes obsolete tests
Provides binaries for macOS (Apple Silicon and Intel), iOS, Linux (CPU, Vulkan, OpenVINO, SYCL), Android, Windows (CPU, CUDA 12/13, Vulkan, OpenVINO, SYCL, ROCm), and openEuler

This release ensures more efficient string handling in template rendering and offers updated binaries across a wide range of hardware architectures and operating systems.

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

Read original

github llama.cpp · 4h ago

llama.cpp b10434 passes reasoning_effort to templates

The llama.cpp project released build b10434, which adds support for passing the `reasoning_effort` parameter through to Jinja templates. This change ensures that OpenAI Chat Completions `reasoning_effort` values are stored and made available during generation.

lab NVIDIA Research · 5h ago · 2 views

Nvidia presents FusionRelight for real-time portrait relighting via hybrid domain knowledge fusion

Researchers from Nvidia have introduced FusionRelight, a method for portrait relighting that combines physically plausible illumination transfer with identity preservation and compact real-time inference. The approach utilizes Hybrid Domain Knowledge Fusion (HDKF), a training framework that distills physics, reflectance, and realism priors from synthetic, One-Light-at-a-Time (OLAT), and in-the-wild data into a student model.

github llama.cpp · 5h ago · 1 view

llama.cpp b10434 passes reasoning_effort to templates

Nvidia presents FusionRelight for real-time portrait relighting via hybrid domain knowledge fusion

llama.cpp b10431 adds recurrent state rollback for SSM models

Qwen releases open weights for Qwen 3.8 27B

llama.cpp b10429 allows accessing /metrics and /slots during llama_decode()