Models & Labs
Llama.cpp Update Fixes Reasoning Budget Handling
llama.cpp Releaseshigh confidence
Why it matters
- →Fixes ensure caller-supplied values are respected, enhancing AI control.
- →Improves precision in AI reasoning processes for chat completions.
- →Validated with unit tests, ensuring reliability and accuracy.
Llama.cpp has released an update that fixes a bug related to reasoning budgets in chat completions. The issue involved the system ignoring per-request values for reasoning_budget_tokens and reasoning_budget_message, defaulting to server-level settings instead. This update ensures that the caller's intended values are used, enhancing the precision of AI reasoning processes. The fix is verified with unit tests, improving the flexibility and accuracy of chat completions.
Read originalMore from llama.cpp Releases
Open Sourcemodels
llama.cpp b9973 Release Expands Platform Support
The b9973 release of llama.cpp focuses on enhancing compatibility across a wide array of systems, though it doesn't introduce major new features. This update is particularly notable for adding ROCm 7.2 support on Ubuntu x64, offering AMD GPU users a viable alternative to NVIDIA's CUDA. The release continues to provide extensive builds for macOS, Linux, Windows, and openEuler, ensuring that developers can deploy llama.cpp in varied environments. While the update lacks groundbreaking innovations, it strengthens llama.cpp's role as a versatile tool for AI inference, accommodating diverse hardware configurations.
llama.cpp Releases
Models & Labsmodels
llama.cpp b9974 Release Fixes CUDA Memory Crash
The latest b9974 release of llama.cpp addresses a critical issue for CUDA users by preventing crashes when querying memory on devices with no free memory. Previously, attempting to check memory availability could lead to a fatal crash if the device was out of memory. The update now assigns zero total/free memory to such devices, ensuring the fit algorithm doesn't attempt to use them, thus avoiding crashes. This change enhances stability for CUDA-enabled builds, especially when users specify '-dev none'. While the update doesn't introduce new features, it significantly improves reliability for developers working with CUDA devices.
More in Models & Labs
Models & Labsmodels
vLLM v0.25.0 Release Enhances Model Execution
The vLLM v0.25.0 release marks a significant step forward in model execution and performance. With Model Runner V2 now the default for dense models, users benefit from enhanced support for real-time embeddings and dynamic speculative decoding. The removal of PagedAttention and improvements to the Transformers backend, including FP8 MoE support, streamline operations and boost speed. New models like LLaVA-OneVision-2 and Unlimited OCR expand the model zoo, offering more options for developers. This release solidifies vLLM's position as a robust platform for AI model deployment, with improved efficiency and expanded capabilities.
vLLM Releases
© The Verge AIModels & Labsmodels
