🤖 The second video in the series about Visual Language Action policies for robotics! This time, I will focus on one of the most notable open-source VLA models, Pi0 from Physical Intelligence. 🎯 What You'll Learn in This Episode: ✅ The model architecture and the overview ✅ Fine-tuning on your own robot ✅ FAST tokenization vs. Flow Matching ✅ LoRA vs Full fine-tuning ✅ Practical demo with LeKiwi and LeRobot 🔥 Coming Up in This Series: - Pi-0.5 - the more advanced version of Pi 0 - SmolVLA from HuggingFace LeRobot - Gr00t N1.5 from NVIDIA - And other related topics 0:00 Intro 3:44 Model Architecture 8:21 Action Expert Details 22:30 Fine-tuning with OpenPI 52:01 Experiments with My Robot 1:14:12 Outro 🎬 Series Playlist: 🔗 Resources & Code: - Pi-0 Blog post: - Pi-0 Paper: - OpenPi Github: - My repository with configs and code for finetuning: - My Dataset: - LeKiwi Robot: - SO-ARM Kit: - LeRobot Library: 🎬 Other related videos: - Introduction to VLAs: - Short intro about SO-ARM: - Big video about SO-ARM and LeRobot: - Intro into LeKiwi: - Multimodality in robotics: - Robot MCP project: 💬 Join the Discussion: What VLA-related topics do you want me to cover next? Drop your ideas below! I read every comment and often feature community suggestions in future videos. 🔔 Don't miss the next episode! Hit the bell icon - this series releases every few weeks with hands-on tutorials, code walkthroughs, and real robot experiments. ⭐ Found this helpful? Like, subscribe, and share with anyone building the future of robotics! #Pi0 #Pi05 #VLA #Robotics #AI #ArtificialIntelligence #MachineLearning #LLM #VLM #transformers #PhysicalIntelligence #HuggingFace #LeRobot #EmbodiedAI #RobotLearning