HierarchicalKV 0.1.0-beta.2

Release Notes

This is the 2nd pre-release of HierarchicalKV!

Changes

• Improve the performance of main APIs and update the benchmark
• Make the dimension of the vector configurable in runtime
• Fix some bugs, for more detail please refer to the commits with '[Fix]' as a prefix

Thanks to our Contributors

@bashimao @mikemckiernan @neuzxy and @rhdong

HierarchicalKV v0.1.0-beta.1

Release Notes

This is the first release of HierarchicalKV!

Features

• Supports training large RecSys models on HBM and host memory at the same time.
• Provides better performance by full bypassing CPUs and reducing the communication workload.
• Implements table-size restraint strategies that are based on LRU or customized strategies. The strategies are implemented by CUDA kernels.
• Operates at a high working-status load factor that is close to 1.0.

Thank You to All Our Contributors

• Fan Yu
• EmmaQiao
• Gems Guo
• Haidong Rong
• LiFan
• Matthias Langer
• Michael McKiernan
• Ranjeet
• 王泽寰
• Zhangyafei

Acknowledgment

We are very grateful to external initial contributors @Zhangyafei and @Lifan for their design, coding, and review work.