Skills with output "Training metrics"

moe-training - Train sparse Mixture of Experts models

Mixture of Experts Sparse Models DeepSpeed Expert Parallelism

★ 3 · Updated 2026-09-29

Train and optimize sparse Mixture of Experts models with DeepSpeed or HuggingFace, including routing, load balancing, expert parallelism, and inference optimization.

⚙ Implement sparse MoE architectures ⚙ Configure top-k routing ⚙ Balance expert utilization

AgentDB Learning Plugins - Create and train reinforcement learning plugins for agents

reinforcement-learning agentdb plugins machine-learning

★ 650 · Updated 2026-09-24

Provides 9 reinforcement learning algorithms to create, train, and manage learning plugins for autonomous agents in AgentDB.

⚙ Create learning plugins ⚙ List plugin templates ⚙ Inspect plugin information

ai-architect-expert - Expert guidance for AI system and MLOps architecture.

ai-architecture mlops system-design ai-infrastructure

★ 2 · Updated 2026-09-23

Provides guidance for AI system design, MLOps architecture, scalable ML infrastructure, and AI platform engineering.

⚙ Design AI system architectures ⚙ Implement MLOps infrastructure ⚙ Build model registries

verl-rl-training - LLM Reinforcement Learning Training with verl

Reinforcement Learning RLHF GRPO PPO

★ 0 · Updated 2026-06-30

Train LLMs with RLHF, GRPO, and PPO using verl library

⚙ train LLMs ⚙ configure algorithm ⚙ prepare dataset

verl-rl-training - verl: LLM RL Training Framework

Reinforcement Learning RLHF GRPO PPO

★ 1 · Updated 2026-06-30

Flexible RL training library for large language models

⚙ train models ⚙ run rollout ⚙ compute reward