Mixture of Experts: Core Concepts and Hands-on Implementation
Chapter 1: Foundations of Mixture of Experts Models
Overview of Sparsely-Gated MoE Architecture
The Gating Network: Formulation and Function
Expert Networks: Specialization and Capacity
Mathematical Formulation of the MoE Layer
Load Balancing and Auxiliary Losses
Challenges in MoE Training: Expert Collapse
Comparison with Dense Model Scaling
Hands-on: Implementing a Basic MoE Layer
Chapter 2: Advanced Routing Mechanisms
Analysis of Top-k Gating and its Variants
Noisy Top-k Gating for Load Balancing
Hash-based Routing for Deterministic Selection
Switch Transformers: Simplified Routing
Soft MoE: Differentiable Routing
Analyzing Routing Decisions and Specialization
Hands-on: Implementing Different Routing Strategies
Chapter 3: Training and Optimization of Large-Scale MoEs
Expert Parallelism for Distributed Training
Combining Model, Data, and Expert Parallelism
Capacity Factor and its Impact on Performance
Techniques for Mitigating Router Z-Loss Instability
Precision and its Effects: BFloat16 Training
Fine-tuning Strategies for Pre-trained MoE Models
Practice: Configuring a Distributed Training Job
Chapter 4: Efficient Inference with MoE Models
Inference Challenges: Memory and Latency
Expert Offloading to CPU or NVMe
Batching Strategies for Sparse Activation
Model Distillation for MoE Compression
Quantization Techniques for MoE Layers
Speculative Decoding with MoE Models
Hands-on: Building an Optimized Inference Pipeline
Chapter 5: Integrating MoE into Modern Architectures
Replacing FFNs with MoE Layers in Transformers
Placement of MoE Layers: Frequency and Location
MoE in Vision Transformers (ViT)
MoE in Multi-modal Models
Architectural Variants and their Properties
Analyzing Parameter vs. FLOPs Trade-offs
Practice: Modifying a Transformer to use MoE
Analyzing Parameter vs. FLOPs Trade-offs
Was this section helpful?
- Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer, Noam Shazeer, Azalia Mirhoseini, Krzysztof Maziarz, Andy Davis, Quoc Le, Geoffrey Hinton, and Jeff Dean, 2017 International Conference on Learning Representations (ICLR) 2017 - Introduces the sparsely-gated Mixture-of-Experts (MoE) layer, demonstrating how it enables a large increase in model capacity without a proportional increase in computational cost per token.
- Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity, William Fedus, Barret Zoph, and Noam Shazeer, 2022 Journal of Machine Learning Research, Vol. 23 (JMLR, Inc. and Microtome Publishing) - Presents Switch Transformers, which simplify MoE layers to one active expert per token (k=1), showcasing their ability to scale models to trillion parameters while addressing communication and load balancing challenges.
- GLaM: Efficient Scaling of Language Models with Mixture-of-Experts, Nan Du, Yanping Huang, Andrew M. Dai, Simon Tong, Dmitry Lepikhin, Yuanzhong Xu, Maxim Krikun, Yanqi Zhou, Adams Wei Yu, Orhan Firat, Barret Zoph, Quoc V. Le, and Zhifeng Chen, 2022 arXiv preprint arXiv:2201.05824 DOI: 10.48550/arXiv.2201.05824 - Details GLaM, an MoE architecture that achieves high performance with significantly fewer training FLOPs compared to dense models of similar quality, emphasizing efficiency and scaling.
- A Survey of Mixture of Experts, Xufeng Lin, Yiming Qian, Yuanyang Liu, Huadong Liu, Xizhen Sun, Jianyang Li, Guanyu Chen, Qingyu Jin, Meng Zhang, and Bo Xu, 2023 arXiv preprint arXiv:2308.14073 (arXiv) DOI: 10.48550/arXiv.2308.14073 - Provides a comprehensive overview of Mixture of Experts models, covering their history, architectural variations, training techniques, and practical considerations like efficiency and load balancing.