moe-training

github:zechenzhangAGI/AI-research-SKILLs · first committed 2025-11-19 · source · badge

Train Mixture of Experts (MoE) models using DeepSpeed or HuggingFace. Use when training large-scale models with limited compute (5× cost reduction vs dense models), implementing sparse architectures like Mixtral 8x7B or DeepSeek-V3, or scaling model capacity without proportional compute increase. Covers MoE architectures, routing mechanisms, load balancing, expert parallelism, and inference optim

Skill Health: not yet scored

Listed, not yet scored. This skill has no public model test yet. It is in the scoring queue; submitted skills go first, then the most-installed.

Version history

5 commits to this file on GitHub (newest first).

2025-11-20 · commit 0a9ded4 · Standardize skill metadata: Add category tags and update authors
2025-11-20 · commit 51fc3a3 · Trigger upload of remaining 32 skills batch 2
2025-11-20 · commit 81c38fe · Upload all remaining 34 skills to Orchestra marketplace
2025-11-19 · commit bd4461f · Standardize YAML frontmatter for all 54 skills
2025-11-19 · commit 66e31d2 · Add 6 Emerging Techniques skills with comprehensive references

Archived snapshots