Skip to content
This repository has been archived by the owner on Nov 17, 2023. It is now read-only.

[FEATURE] Add interleaved batch_dot oneDNN fuses for new GluonNLP models #20312

Merged
merged 8 commits into from
Jun 29, 2021
Merged

[FEATURE] Add interleaved batch_dot oneDNN fuses for new GluonNLP models #20312

merged 8 commits into from
Jun 29, 2021
+1,706 −2

Commits on Jun 29, 2021

  1. Add self-attention fuse for MKLDNN backend 

    Float self attention fuse

WIP

Add transpose - reshape

Add self attention fuse with oneDNN support
    @bgawrych
    bgawrych committed Jun 29, 2021
    Configuration menu
    Copy the full SHA
    5668d46 View commit details
    Browse the repository at this point in the history

  2. Review

    @bgawrych
    bgawrych committed Jun 29, 2021
    Configuration menu
    Copy the full SHA
    ca90e83 View commit details
    Browse the repository at this point in the history

  3. Fix sanity

    @bgawrych
    bgawrych committed Jun 29, 2021
    Configuration menu
    Copy the full SHA
    6b551d6 View commit details
    Browse the repository at this point in the history

  4. Make ConvertWeightBias2MKLDNN inline function

    @bgawrych
    bgawrych committed Jun 29, 2021
    Configuration menu
    Copy the full SHA
    4d16c43 View commit details
    Browse the repository at this point in the history

  5. Fix CI compilation errors

    @bgawrych
    bgawrych committed Jun 29, 2021
    Configuration menu
    Copy the full SHA
    d53e14e View commit details
    Browse the repository at this point in the history

  6. Add new fp32 ops names to amp list

    @bgawrych
    bgawrych committed Jun 29, 2021
    Configuration menu
    Copy the full SHA
    e474df8 View commit details
    Browse the repository at this point in the history

  7. Remove double semicolon

    @bgawrych
    bgawrych committed Jun 29, 2021
    Configuration menu
    Copy the full SHA
    70d3c89 View commit details
    Browse the repository at this point in the history

  8. Switch to forward interface

    @bgawrych
    bgawrych committed Jun 29, 2021
    Configuration menu
    Copy the full SHA
    88401a0 View commit details
    Browse the repository at this point in the history