Reinforcement learning

Agent Apprenticeship is an open infrastructure platform for creating learning loops where AI agents complete real-world tasks, receive feedback from...

ProtoMotions3 is a GPU-accelerated simulation and learning framework specifically designed for training physically simulated digital humans and...

Nemotron is NVIDIA's developer hub for training, fine-tuning, and deploying the Nemotron family of open models optimized for agentic AI workflows. It...

RLinf is an open-source reinforcement learning infrastructure designed specifically for embodied AI and agentic systems, enabling RL fine-tuning of...

An open-source, unconventional textbook covering mathematics, computer science, and artificial intelligence fundamentals, designed for practitioners...

NeMo Gym is a library for evaluating and improving AI models and agents in stateful environments (e.g., code execution, tool use), designed...

AReaL is a large-scale asynchronous reinforcement learning infrastructure designed to train reasoning and agentic models for LLM-based applications....

ART is an open-source reinforcement learning framework that trains multi-step LLM-based agents using GRPO, enabling models like Qwen, Llama, and...

This repository is the official code companion to a Manning book that teaches how to build a reasoning LLM from scratch using PyTorch, starting with...

A comprehensive, open-source AI engineering curriculum spanning 503 lessons across 20 phases that teaches AI fundamentals from first principles...

mjlab is a GPU-accelerated reinforcement learning and robotics simulation framework combining Isaac Lab's manager API with MuJoCo-Warp. It...

Mooncake is a specialized KVCache-disaggregated serving platform for large language model inference, designed to handle high-throughput request...

A collection of 18 reinforcement learning algorithms implemented from scratch in Jupyter Notebooks, designed explicitly for educational understanding...

robosuite is a modular simulation framework powered by MuJoCo for robot learning research, offering benchmark environments, diverse robot...

SGLang is a high-performance inference serving framework optimized for large language models and multimodal models, with support for speculative...

TorchRL is a PyTorch-native reinforcement learning library providing modular, composable components for building RL systems including environments,...

This is a comprehensive code repository and educational resource for the third edition of 'Machine Learning for Trading,' covering the complete...

A curated collection of 100+ original architecture diagrams and visual maps explaining LLM, reinforcement learning, and vision-language model...

SANA is a specialized research framework for efficient high-resolution image and video generation using linear diffusion transformers, developed by...

MetaClaw is an AI agent framework that learns and adapts from conversations through meta-learning and continual fine-tuning, requiring no GPU...

A curated collection of AI learning resources—books, courses, papers, and tools—organized by topic and skill level. It serves as a structured entry...

OSWorld is a NeurIPS 2024 benchmark for evaluating multimodal AI agents on open-ended real computer tasks, supporting multiple virtualization...

OpenRLHF is a production-ready framework for reinforcement learning from human feedback (RLHF) built on Ray and vLLM, enabling scalable training of...

rLLM is an open-source framework for training language agents using reinforcement learning, designed to work with any harness (Claude, Codex, custom...

EnvPool is a C++-based batched environment execution engine that achieves 1M+ FPS for RL training by vectorizing environment simulation across dozens...

RLtools is a C++ deep reinforcement learning library optimized for training speed, particularly on embedded and resource-constrained hardware like M1...

Unsloth Studio is a local web UI and training framework for running, fine-tuning, and deploying open-source LLMs (Llama, Gemma, Qwen, DeepSeek, etc.)...

Judgeval is an open-source Python SDK for agent monitoring and continuous improvement that provides OpenTelemetry-based tracing, agent-judge...

XuanCe is a comprehensive deep reinforcement learning library implementing 20+ DRL algorithms (DQN, PPO, MAPPO, QMIX, etc.) with support for multiple...

IR-SIM is a lightweight Python-based robot simulator designed for navigation, control, and multi-agent learning scenarios. It excels as a rapid...