Skip to content
@RL-Align

RL-Align

Extreme infrastructure and specialized custom kernels for large-scale reinforcement learning.
README.md

Welcome to RL-Align

Extreme infrastructure and specialized custom kernels for large-scale reinforcement learning.

We are an open-source collective dedicated to breaking the memory and compute walls in RLHF and GRPO training pipelines. Our mission is to provide end-to-end, high-performance CUDA/Triton operators that seamlessly integrate into your existing training and rollout frameworks.

Core Projects

  • RL-Kernel: Our flagship project. A high-performance, memory-efficient kernel library featuring prefix_shared_attention and fused_logp for GRPO workloads.

Join the Community

Whether you are a kernel hacker, an AI Infra engineer, or an LLM researcher, you are welcome here!

Pinned Loading

  1. RL-Kernel RL-Kernel Public

    Modern RL Post-training Infrastructure: Optimized for NVIDIA/AMD GPUs with a focus on vLLM and DeepSpeed integration, CUDA/ROCm/Triton kernels, and transparent hardware-aware scaling.

    Python 92 16

Repositories

Showing 2 of 2 repositories

People

This organization has no public members. You must be a member to see who’s a part of this organization.

Top languages

Loading…

Most used topics

Loading…