google-deepmind/rlax

Health v2 · maintenance only

Adoption not part of the score

1439 stars · 106 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

RLax is a JAX-based library of building blocks for implementing reinforcement learning agents, providing mathematical operations like value functions, return distributions, and policy gradients. It is not a complete algorithm framework but composable functions that can be JIT-compiled for CPU, GPU, and TPU.

Use cases

• implement reinforcement learning agents in JAX
• compute TD losses and Q-learning updates
• implement distributional value functions
• build policy gradient agents for discrete and continuous actions
• learn general value functions in JAX
• run RL experiments on GPU or TPU with jit compilation

When to choose

• you are building custom RL agents in JAX and need well-tested mathematical building blocks
• you want composable RL operations rather than monolithic agent implementations
• you need hardware-accelerated RL with jax.jit on CPU, GPU, or TPU

When to avoid

• you want a complete out-of-the-box RL algorithm or training framework
• you work in PyTorch or TensorFlow rather than the JAX ecosystem
• you need environment implementations or full training pipelines

Facets

library · maturity active

machine-learning reinforcement-learning reinforcement-learning machine-learning deep-learning python jax reinforcement-learning deepmind value-functions policy-gradients gpu tpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="google-deepmind/rlax")