Maria1Petrova
Maria1Petrova
Is the source code confusing to understand? Is an optimization hard to follow? We invite you to share your improvements that will reduce the friction for you and others in...
Is something not working the way you expected? Did you find an outright failure? Please bring these issues to our attention, and contribute a fix if you can!
We seek contributions that would allow oneCCL to support the ARM architectures in all computing environments.
We are seeking to expand support outside the Data Center and to enable oneCCL for any GPUs of interest on client systems.
Often times, problems manifest when running over many hours that do not easily manifest when running shorter duration workloads. We seek improvements that help improve the robustness of oneCCL over...
oneCCL exists in order to facilitate high performance communications through collective operations spanning high node count CPU and/or GPU clusters. We seek enhancements to maximize the throughput and minimize latency...