单卡即可微调大模型！内存占用仅1/8，性能依然拉满 | ICML 2025
"混合专家" 相关文章
原生多模态模型的标度律：重新思考架构选择与训练效率
原生多模态模型的标度律：重新思考架构选择与训练效率
苹果提出原生多模态Scaling Law！早融合+MoE，性能飙升秘密武器
探秘Transformer系列之（21）--- MoE
SYMBOLIC-MOE: Mixture-of-Experts MoE Framework for Adaptive Instance-Level Mixing of Pre-Trained LLM Experts
ByteDance AI Introduces Doubao-1.5-Pro Language Model with a ‘Deep Thinking’ Mode and Matches GPT 4o and Claude 3.5 Sonnet Benchmarks at 50x Cheaper
Monet: Mixture of Monosemantic Experts for Transformers Explained
Mixture-of-Denoising Experts (MoDE): A Novel Generalist MoE-based Diffusion Policy
DeekSeek v3: The Six Million Dollar Model
50张图，直观理解混合专家（MoE）大模型
Jurgen、曼宁等大佬新作：MoE重塑6年前的Universal Transformer，高效升级
XVERSE-MoE-A36B Released by XVERSE Technology: A Revolutionary Multilingual AI Model Setting New Standards in Mixture-of-Experts Architecture and Large-Scale Language Processing
Mixture-of-Experts (MoE) Architectures: Transforming Artificial Intelligence AI with Open-Source Frameworks
Wolf: A Mixture-of-Experts Video Captioning Framework that Outperforms GPT-4V and Gemini-Pro-1.5 in General Scenes, Autonomous Driving, and Robotics Videos
算法、系统和应用，三个视角全面读懂混合专家（MoE）
DeepSeek AI Researchers Propose Expert-Specialized Fine-Tuning, or ESFT to Reduce Memory by up to 90% and Time by up to 30%
NVIDIA经济学：云服务商每花1美元买我的GPU 就能赚7美元！
Two AI Releases SUTRA: A Multilingual AI Model Improving Language Processing in Over 30 Languages for South Asian Markets
Training on a Dime: MEFT Achieves Performance Parity with Reduced Memory Footprint in LLM Fine-Tuning
