Reinforcement Learning with TorchRL and TensoDict - NeurIPS Hacker Cup AI

Dive into an in-depth live session exploring TorchRL and TensoDict, powerful tools for reinforcement learning in PyTorch. Learn from Vincent Moens of the TorchRL team at Meta as he breaks down the structure and key features of TorchRL, including its application in fine-tuning models through reinforcement learning from human feedback (RLHF). Gain insights into practical applications, efficient data handling, and how TensoDict streamlines RL workflows. Explore topics such as vectorized maps, multiagent systems, parameter serialization, and model execution. Perfect for developers and researchers looking to enhance their skills in advanced reinforcement learning techniques and leverage cutting-edge PyTorch tools.

Introduction
Reinforcement learning
Step by step
Open Source Repose
Torl
Vectorized Maps
Multiagent
PO
Tenso
Tenso representation
Name tensors
Data representation
Parameter serialization
Vectorized Mass
Model Execution
Tensor Class
Hot Takes