LEARN BY BUILDING
Learn reinforcement learning, language models and computer vision through open courses and hands-on projects.
Reinforcement learning · Courses & experiments
Modern reinforcement learning: from fundamentals to LLMs and agents
Combine concepts, mathematical derivations and coding experiments to learn value functions, DQN, policy gradients and PPO, then explore RLHF, DPO, GRPO and Agentic RL.
Start hereRead the introduction and setup instructions, then complete the chapter experiments. The course is still developing; start with completed chapters.
Language models · Hands-on code
Understand language-model training through a complete codebase
Andrej Karpathy's open-source language-model training project connects tokenization, pretraining, fine-tuning, evaluation and inference in concise, editable code for studying the full training process.
Start hereBegin with the README, then follow the training scripts. Choose a workflow suited to your hardware.
Language models · Efficient fine-tuning
From dataset preparation to model fine-tuning
A fine-tuning framework for language and vision-language models, with supervised fine-tuning, LoRA and QLoRA workflows through a command-line interface or visual interface.
Start hereFollow the official quickstart to prepare data and fine-tune a small model with LoRA, then explore evaluation and inference.
Reinforcement learning · LLM post-training
Put PPO and GRPO into practice
A reinforcement learning framework for LLM post-training, with PPO and GRPO workflows covering data preparation, reward functions, rollout generation and distributed training.
Start hereStart with installation and the quickstart. Learn how data, reward functions and training configuration fit together before running an official example.
Computer vision · Text recognition
From text recognition to document parsing
A text recognition and document parsing toolkit in the PaddlePaddle ecosystem, covering text detection, recognition, tables and document layouts for hands-on work with images and documents.
Start hereRecognize text in a single image first and inspect the detection and recognition results, then explore document structure parsing and application integration.
Computer vision · Object detection
Learn the YOLO family with Ultralytics
Explore YOLO models through the official Ultralytics implementation, from object detection, instance segmentation and pose estimation to tracking, with training, validation, prediction and model export workflows.
Start hereRun predictions with a pretrained model, then train a detector on your own annotated data and explore validation and export.
More tutorials to come.
Resources are maintained by their original authors. Follow each link to learn on the source site.
