Skip to content

MindSpeed-RL

MindSpeed RL is a reinforcement learning acceleration framework based on the Ascend ecosystem, aiming to provide end-to-end RL training and inference solutions for Huawei Ascend chip ecosystem partners, supporting core acceleration capabilities such as large-scale Ascend cluster training and inference deployment, multi-model asynchronous流水调度, training and inference heterogeneous split communication, etc.

The guidance for using MindSpeed RL with SwanLab is available at: https://gitcode.com/Ascend/MindSpeed-RL/blob/master/docs/features/logging_swanlab.md