How to train a MoE-LLM in 2026
Practical presentation on expert routing, GPU parallelism, profiling, and fine-tuning mixture-of-experts language models.
Presentation ↗I hike, speedrun games, go to concerts and collect stickers and figures.
Surprise meDrag a sticker or use the arrow keys to turn it. Home restores the front view.
I met my lovely girlfriend Arina in Germany.
Two days, 25 km and 2,300 m up to the 2,962 m summit.
Finally, a qualification with immediate practical use.
Practical presentation on expert routing, GPU parallelism, profiling, and fine-tuning mixture-of-experts language models.
Presentation ↗Presentation of the thesis method, experiments, efficiency tradeoffs, and limitations on multi-step software engineering.
Presentation ↗Seven sections. Scroll to split.
In order. Green: ahead. Red: behind. Gold: best split. First run: random ghost.