检索标题、摘要与正文，把分散的信息连起来。
本文为Hugging Face与Amazon联合发布的技术博客系列首篇，系统介绍如何在AWS Trainium AI加速芯片上使用TorchNeuron进行PyTorch模型推理性能分析，涵盖Neuron后端集成原理、torch.profiler配置方法、NeuronConfig与NeuronProfiler工具使用、Neuron Explorer可视化追踪，并以RMSNorm为例对比三种实现（朴素PyTorch、torch.rms_norm内置融合、torch.compile自动融合）的设备耗时差异。
i14推出30.28亿参数的Laplace 3B模型，采用门控线性注意力（GLA）替代Softmax注意力，其推理时每序列递归状态恒为29.5 MB，与上下文长度无关；该模型在单台AWS trn2.48xlarge（16颗Trainium2芯片）上完成训练与服务验证，训练吞吐达12万token/秒，服务吞吐11544 output token/秒（1280并发序列），所有性能数据均为i14与AWS GenAIIC团队内部实测结果。
SF Compute由Evan Conrad创立，聚焦于实现GPU算力的可编程、物理交割式交易；其不拥有数据中心，而是通过全栈技术协同第三方设施共建‘主权算力云’，强调市场中立性与基础设施可控性，已实现数亿美元交易额并落地InfiniBand动态切分等关键能力。
为应对生成式AI爆发带来的高带宽内存芯片需求激增，三星在韩国平泽建设占地相当于纽约中央公园的半导体制造集群，目前已投运三座晶圆厂，第四座即将投产，第五座在2024年因需求下滑暂停后已重启建设；现场呈现高强度、高安全标准的建设节奏与劳动力生态。
Hugging Face 发布 TorchTPU，为 PyTorch 提供原生 TPU 后端支持，使 Transformers、Diffusers 等库可直接在 TPU 上运行，无需依赖 torch_xla 或迁移到 JAX。
Hugging Face正式发布AnyLanguageModel 1.0，提供与Apple Foundation Models API完全兼容的开源Swift框架，支持本地（MLX、llama.cpp、Core ML）及托管（OpenAI、Anthropic、Gemini、Hugging Face Inference Providers等）模型无缝切换，并新增图像输入、工具调用、引导式生成、推理块自动适配等特性。
Hugging Face联合AWS推出TorchNeuron——PyTorch原生Trainium后端，自2026年10月20日起开放公测。该技术使Transformers、Diffusers等主流库无需修改即可直接运行于Trainium芯片，显著提升文本生成与图像生成模型的易用性与兼容性。
Differential acceleration of cyber, math, and AI
译文排队中，当前为原文 · enGPU 管理:为什么摇摆 GPU 是新地面飞机
Figure 1: CUDA-to-MLX optimization translation map. CUDA optimization knowledge can be translated into architecture-native MLX strategies rather than copied instruction-for-instruction. We face a new epoch in computing. Hardware is changing rapidly — not just faster GPUs, but a growing range of chips from different vendors, each with its own architecture and often tailored to specific AI workloads. Software is changing just as fast, and AI coding tools now generate in minutes what took months of effort a few years ago. With so much of computing now centered on AI, GPU kernels are a crucial component of its success. These are the low-level programs that run inside the GPU, and writing efficient ones is far from obvious — it takes years of expertise to get right. Transferring a kernel from o
译文排队中，当前为原文 · en... government of the people, by the people, for the people ... — Abraham Lincoln, Gettysburg Address (1863) The cost of AI is dropping rapidly. GPT-4-class capabilities cost roughly $30 per million tokens in early 2023; today the same runs under $1, and some providers are pushing costs below $0.10. Across benchmarks, inference prices have fallen between 9x and 900x per year, with a median decline near 50x. Even frontier models are getting dramatically cheaper each generation, with open-source models following closely behind. And crucially, even if “Nobel-Prize-winning genius-level” intelligence isn’t here yet, the intelligence that suffices for the vast majority of knowledge work is here today, and getting cheaper by the month. At this rate, we are soon entering the era of virtually free
译文排队中，当前为原文 · enIs this the beginning of a new world?
译文排队中，当前为原文 · enWhat eras bookend our interregnum?
译文排队中，当前为原文 · en