PufferLib Docs

These docs will get you started. Join the Discord to get help and report bugs. If you're new to RL, building and contributing a new env is the best way to learn, and we review PRs live. The PufferLib 4.0 native backend is ~1500 lines of Python and ~5000 lines of CUDA C. A PyTorch backend is provided as an additional ~1000 lines of Python for quick prototyping and as a fallback. Fork the project and edit directly. There is no prebuilt package blackboxing functionality. Everything is written as simply and transparently as possible. If you're new to low-level dev, it's much easier than you think. Give the environment tutorial a try! Memory Management: Tensors in PufferLib are just structs with a shape and a data pointer. Every tensor registers its size with an allocator at init time. After all tensors are registered, the allocator sums up the sizes and does a single allocation of continuous memory. There are separate allocators for weights, gradients, and activations. No tensors are creat

saved by

• Keane Moraes