LLM: обучение и использование. 5. MoE, Deep Seek, Qwen3

  1. 00:05Mixture of Experts (MoE)
  2. 30:14Deep Seek
  3. 01:00:15Reasoning модели
  4. 01:03:58GRPO vs PPO
  5. 01:11:29Qwen3
  6. 01:23:57Discussion