新发现
red-weasel/machx-inference-engine v0.2.35 — Flash-Next 链
基于 C++/SYCL 的 Intel Arc 推理引擎 machx-inference-engine 在 v0.2.35 中加入了把 Flash-Next 拆分到多张卡上运行的 N 阶段链,并修复了社区 GGUF 的读取错误。
基于 C++/SYCL 的 Intel Arc 推理引擎 machx-inference-engine 在 v0.2.35 中加入了把 Flash-Next 拆分到多张卡上运行的 N 阶段链,并修复了社区 GGUF 的读取错误。
GitHub 项目说明: C++/SYCL local LLM inference for Intel Arc: Qwen3.8, DeepSeek-V4.1, MiMo-V2.6, GLM-5.3 and LTX-2.5 text-to-video on Arc Pro B70 cards. XMX kernels, quantized models, multi-GPU.
在 GitHub 查看 ↗基于 C++/SYCL 的 Intel Arc 推理引擎 machx-inference-engine 在 v0.2.35 中加入了把 Flash-Next 拆分到多张卡上运行的 N 阶段链,并修复了社区 GGUF 的读取错误。
没有符合条件的动态。