Moe Training

This walks you through training Mixture of Experts models with DeepSpeed or HuggingFace, covering everything from basic MoE layers to production configurations like Mixtral 8x7B. You get working code for routing mechanisms (top-1, top-2, expert choice), load balancing losses to prevent expert collapse, and expert parallelism configs for distributing 128 experts across multiple GPUs. The real value is in the details: capacity factors, auxiliary loss coefficients, router z-loss for stability. Use this when you need to train larger models on limited compute, the claimed 5x cost reduction versus dense models is legit because you only activate a fraction of parameters per token. The DeepSpeed configurations are production ready, not toy examples.

Install to Claude Code

npx -y skills add orchestra-research/ai-research-skills --skill moe-training --agent claude-code