Switch Transformer Contents MoE: Mixture of Experts MoE: Mixture of Experts sparsely-activated model: 为每个传入的样本选择不同的参数。拥有庞大的参数，然而计算代价不变。 Switch Transformer: 简化MoE路由算法，减少通讯和计算开销。 Share on: Twitter ❄ Facebook ❄ Email Comments Comments Related Posts SELF-INSTRUCT: Aligning Language Model with Self Generated Instructions LLM Algorithms Universal Chart Structural Multimodal Generation and Extraction MinHash: Document-level Deduplication GGUF Model Published Jan 21, 2025 Category NLP Tags LLM 8 MoE 1 NLP 36 Contact
