There is a newer version of the record available.

Published October 21, 2022 | Version v0.7.0
Software Open

HybridBackend

Description

Objectives:

  1. Memory-efficient loading of categorical data
  2. GPU-efficient orchestration of embedding layers
  3. Communication-efficient training and evaluation at scale
  4. Easy to use with existing AI workflows

Features:

  1. Performance

    • Support of automatic embedding fusion on PAI DLC / PAI DSW
    • Support of data transfer prefetching
  2. Usability

    • Support of embedding_lookup_* API
    • Support of Keras Model API
    • Support direct pip install via Pypi

Notes

Please cite HybridBackend in your publications if it helps

Files

alibaba/HybridBackend-v0.7.0.zip

Files (1.3 MB)

Name Size Download all
md5:1af6b44d5e84daaf8d6ff76fb49667cb
1.3 MB Preview Download

Additional details

Related works