大模型
x.com ·
@AI at Meta (@AIatMeta)
Muse Spark Open 优化本地部署:支持 Apple M5 Ultra、MLX 与主流推理引擎
Meta 表示 Muse Spark Open 针对两台 Apple M5 Ultra Mac Studio 与最新 MLX 框架做了优化,同时可在 SGLang、vLLM、TensorRT-LLM、Hugging Face Transformers、llama.cpp 和 Unsloth 等推理引擎上运行,并针对 AMD、Apple、Intel、NVIDIA 芯片优化。开发者可用 PyTorch(含 TorchTitan)进行微调。帖中未提供性能或吞吐数据。
AI推荐理由 同一发布系列的补充信息,具体说明 Muse Spark Open 可在两台 Apple M5 Ultra Mac Studio 上通过 MLX 框架运行,并支持 SGLang、vLLM、TensorRT-LLM、Hugging Face Transformers、llama.cpp、Unsloth 等推理引擎,覆盖 AMD、Apple、Intel、NVIDIA 芯片,且可用 PyTorch 与 TorchTitan 微调。对本地部署与硬件适配有实际参考价值;但同样缺少性能数据与量化指标,无法验证优化效果。
78
浏览 83,169
点赞 696
评论 18
转发 35
正文
To help make advanced models more accessible, Muse Spark Open is optimized to run on two Apple M5 Ultra Mac Studios with the latest MLX framework.
It also runs on leading inference engines like SGLang, vLLM, TensorRT-LLM, Hugging Face Transformers, llama.cpp, and Unsloth with optimizations for AMD, Apple, Intel, and NVIDIA silicon.
Developers can further customize the model with PyTorch, including TorchTitan for fine-tuning.