github llama.cpp · 1h ago · inference

llama.cpp b10604 adds Deepseek 4 DSV4 support and macOS/iOS binaries

from English

The llama.cpp project has released version b10604, introducing support for the Deepseek 4 (DSV4) model format alongside new precompiled binaries for macOS and iOS.

Adds Deepseek 4 (DSV4) sm tensor support with coarser granularity for head splits and fixes for dspark.
Enables model saving for DSV4 and allows dflash to return on specific devices.
Includes shared expert delayed allreduce functionality.
Provides macOS Apple Silicon binaries, iOS XCFrameworks, and updated binaries for Linux, Windows, Android, and openEuler across CPU, GPU, and specialized backends.

This release enables users to run Deepseek 4 models using llama.cpp and provides updated native builds for Apple devices.

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

Read original

github llama.cpp · 14h ago · 5 views

llama.cpp b10598 uses pillow-accurate resize algo and corrects resize_algo for all models

The llama.cpp project released build b10598, which updates the mtmd module to use a pillow-accurate resize algorithm and corrects the resize_algo for all models. This release also includes speed optimizations.

github llama.cpp · 21h ago · 3 views

llama.cpp b10593 fixes DeepseekV4 rollback and model loading

The llama.cpp project has released build b10593, which addresses critical issues with the DeepseekV4 model architecture. The update specifically resolves a rollback failure that occurred during multi-sequence processing.

github llama.cpp · 2d ago · 3 views

llama.cpp b10584 fixes draft context sizing to prevent server 500 errors

The llama.cpp project released build b10584, addressing a critical issue where the draft context size did not match the target context, causing server requests to fail with 500 errors.

github llama.cpp · 2d ago · 8 views

llama.cpp b10568 uses ggml_rope_set_offset() for DeepSeek 2

The llama.cpp project released build b10568, which updates the model handling to use the `ggml_rope_set_offset()` function. This change is partially applied to support DeepSeek 2.

media Hugging Face Forums · 3d ago · 6 views

Celestium Engine adds VRAM optimizations and CPU fallback for SDXL

The Celestium Engine has introduced several upgrades to improve SDXL performance on low-VRAM hardware. These changes focus on memory management, stability, and fallback mechanisms to ensure consistent generation even on constrained systems.