PulseAugur
中
实时 15:23:20
English(EN) Sizing a looped LLM: SenseNova-Loop-9B stores 32 layers, runs 56, and fits a 20 GB card

商汤科技SenseNova-Loop-9B模型以其独特的架构重新定义参数尺寸

商汤科技发布了SenseNova-Loop-9B-Preview,这是一个开放权重模型,在参数尺寸方面提出了独特的挑战。虽然它拥有约90亿参数,但其循环架构每个token需要56层传递,使其推理参数接近146亿。该模型需要llama.cpp、vLLM或SGLang等专业服务引擎,因为标准实现尚不支持其循环结构。测试表明,CPU性能可能成为瓶颈,禁用循环以加快处理速度会导致输出不正确,这表明循环是其训练功能不可或缺的一部分。 AI

影响 引入了一种新的模型架构,挑战了传统的参数尺寸调整,并需要专门的推理引擎。

排序理由 发布了一个具有新颖架构特性的开放权重模型。[lever_c_demoted from research: ic=1 ai=1.0]

在 dev.to — LLM tag 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。 我们如何撰写摘要 →

商汤科技SenseNova-Loop-9B模型以其独特的架构重新定义参数尺寸

本文如何被排名

Signal score
31 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
发布了一个具有新颖架构特性的开放权重模型。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
model release, infra
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Breaking (< 6h)
Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

完整方法见我们的编辑标准。

报道来源 [1]

  1. dev.to — LLM tag TIER_1 English(EN) · InferenceKey ·

    为循环LLM确定尺寸:SenseNova-Loop-9B存储32层,运行56层,并适配20GB显卡

    <p>The parameter count is the number most of us use to size a self-hosted model. SenseNova-Loop-9B-Preview, which SenseTime released with open weights under Apache 2.0, makes that shortcut misleading twice. It takes the memory of a 9B, and its authors count the work per token lik…