GGML
PulseAugur coverage of GGML — every cluster mentioning GGML across labs, papers, and developer communities, ranked by signal.
1 天有情绪数据
-
开发者使用 Vulkan 在 50 美元的 AMD RX 580 GPU 上运行 LLMs
一位开发者展示了如何在拥有 8GB 显存的旧款 AMD RX 580 GPU 上运行大型语言模型和图像生成软件，这在以前被认为是不可能的壮举。通过利用为 ggml 项目（支持 llama.cpp 和 stable-diffusion.cpp 等工具）提供的 Vulkan 后端，该开发者实现了比仅使用 CPU 处理高出 3-4 倍的性能。这种方法绕过了对 CUDA、ROCm 或 DirectML 的需求，证明了现代 AI 任务可以在更普…
-
Google's Gemma 4 adds MTP for faster local inference, VibeVoice ported to C++, Ollama gets desktop layer
Google has released Gemma 4 with Multi-Token Prediction (MTP), a feature that allows the model to predict multiple tokens simultaneously, significantly speeding up local inference. Additionally, a C++ port of Microsoft'…
-
Ollama v0.6.8 and OpenClaw 2026.5.3 release with speedups and fixes
Ollama has released version 0.6.8, introducing performance enhancements for the Qwen 3 MoE model on both NVIDIA and AMD hardware. This update also addresses several issues, including problems with GGML assertions, image…
-
Hugging Face integrates GGML and llama.cpp for local AI
Hugging Face has announced that GGML and llama.cpp are joining the platform. This integration aims to foster the continued development and long-term progress of local AI initiatives. The move is expected to benefit the …
-
George Hotz的tiny corp推出1.5万美元AI电脑和基于RISC的tinygrad框架
George Hotz的公司tiny corp发布了tinybox，一款售价1.5万美元的个人AI电脑，专为本地模型训练和推理设计。tinybox拥有738 FP16 TFLOPS和144 GB GPU内存，能够直接运行65B LLaMA模型。Hotz通过tinygrad框架采用RISC理念以提高效率，并避免图灵完备内核，旨在通过专注于开发者体验和优化现成硬件来与NVIDIA等成熟厂商竞争。