➡️
继续阅读
-
Presentation: From S3 to GPU in One Copy: Rethinking Data Loading for ML Training
Onur Satici explains how Vortex, an open-source columnar file format under th...
-
CD sales are booming as physical media continues its resurgence
According to the Recording Industry Association of America (RIAA), CD sales e...
-
Kimi K3 模型结构(7):结构决定系统
本文总结Kimi K3模型结构对训练与推理系统的影响。KDA采用三种执行形态(FlashKDA、设备内上下文并行、状态回滚)处理固定大小状态;上下文并行通...
-
Kimi K3 模型结构(6):输入端与零件,MoonViT-V2、MTP、Per-Head Muon
Kimi K3模型采用从零训练的MoonViT-V2视觉编码器,27层、patch14结构,支持高分辨率图像token化;MTP层预训练后微调为EAGLE...
-
EP224:MCP与RAG及AI代理的对比
An AI agent is kind of an AI system where the agent performs the task autonom...
-
Kimi K3 模型结构(5):宽度维度,Stable LatentMoE
Kimi K3模型采用Stable LatentMoE作为FFN,基于NVIDIA LatentMoE改进,将输入压缩至半宽latent空间,增加RMSN...