This repository has been archived by the owner on Nov 17, 2023.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

Commits

Commits on Aug 12, 2022

Add support for bool data type for condition in where operator (#21103 )
bgawrych
committedAug 12, 2022

Commits on Aug 4, 2022

Add size threshold for few oneDNN operators (#21106 )

committedAug 4, 2022

Commits on Jul 19, 2022

Reduce overhead in sg_onednn_fully_connected for floats (#21092 )
bgawrych
committedJul 19, 2022

Commits on Jul 6, 2022

oneDNN FullyConnected weight caching & refactor (#21047 )
bgawrych
committedJul 6, 2022

Commits on Jun 28, 2022

[FEATURE] Add tanh approximation for GeLU activation (#21034 )
bgawrych
committedJun 28, 2022

Commits on Jun 14, 2022

fix linkcheck (#21057 )
bgawrych
committedJun 14, 2022

Commits on Jun 7, 2022

broadcast_like CPU optimization (#21004 )

bgawrych
and
Bartlomiej Gawrych
committedJun 7, 2022

Commits on May 19, 2022

fix transformer optimization for gpt-2 (#21007 )
bgawrych
committedMay 19, 2022

Commits on Apr 6, 2022

Restore quantized RNN to master (#20952 )

bgawrych
and
Bartlomiej Gawrych
committedApr 6, 2022

Commits on Mar 4, 2022

Add oneDNN support for "where" operator (#20862 )

bgawrych
and
bartekkuncer
committedMar 4, 2022

Commits on Feb 21, 2022

Reduce after quantization memory usage (#20894 )
bgawrych
committedFeb 21, 2022

Commits on Feb 17, 2022

[website] Add CPU quantization tutorial (#20856 )

committedFeb 17, 2022

Commits on Feb 14, 2022

Improve MaskedSoftmax by oneDNN (#20853 )
bgawrych
committedFeb 14, 2022

Commits on Feb 9, 2022

Fix data-api links (#20879 )
bgawrych
committedFeb 9, 2022

Commits on Feb 5, 2022

Add quantization API doc and oneDNN to migration guide (#20813 )

committedFeb 5, 2022
Fix data-api links (#20867 )

bgawrych
and
Bartlomiej Gawrych
committedFeb 5, 2022

Commits on Jan 19, 2022

Optimize 'take' operator for CPU (#20745 )

committedJan 19, 2022

Commits on Jan 13, 2022

Improve split operator by oneDNN reorder primitive (#20757 )
bgawrych
committedJan 13, 2022

Commits on Jan 6, 2022

Fix identity fuse for oneDNN (#20767 )

bgawrych
and
Bartlomiej Gawrych
committedJan 6, 2022

Commits on Jan 5, 2022

Fix oneDNN fallback for concat with scalar (#20772 )
bgawrych
committedJan 5, 2022

Commits on Dec 24, 2021

Avoid redundant memcpy when reorder not in-place (#20746 )

bgawrych
and
Bartlomiej Gawrych
committedDec 24, 2021
remove broken links (#20793 )

bgawrych
and
Bartlomiej Gawrych
committedDec 24, 2021

Commits on Dec 14, 2021

Fix link check (#20773 )

bgawrych
and
Bartlomiej Gawrych
committedDec 14, 2021

Commits on Dec 2, 2021

Fix test_numpy_op tests & lacking asserts (#20756 )

bgawrych
and
Bartlomiej Gawrych
committedDec 2, 2021
Remove identity operators from oneDNN optimized graph (#20712 )

bgawrych
and
Bartlomiej Gawrych
committedDec 2, 2021

Commits on Dec 1, 2021

Fix sanity CI (#20763 )

bgawrych
and
Bartlomiej Gawrych
committedDec 1, 2021

Commits on Nov 26, 2021

Add oneDNN support for reduce operators (#20669 )

bgawrych
and
Bartlomiej Gawrych
committedNov 26, 2021

Commits on Nov 22, 2021

Improve stack operator performance by oneDNN (#20621 )

bgawrych
and
Bartlomiej Gawrych
committedNov 22, 2021

Commits on Nov 4, 2021

Optimize preparation of selfattn operators (#20682 )
bgawrych
committedNov 4, 2021

Commits on Oct 30, 2021

Add quantized batch_dot (#20680 )
bgawrych
committedOct 30, 2021

Commits on Oct 19, 2021

[Performance] Add oneDNN support for temperature parameter in Softmax (#20567 )
bgawrych
committedOct 19, 2021

Commits on Sep 8, 2021

Add check for MKL version selection (#20562 )
bgawrych
committedSep 8, 2021

Commits on Sep 6, 2021

[Backport] Enabling BRGEMM FullyConnected based on shapes (#20568 )
bgawrych
committedSep 6, 2021

Commits on Jul 30, 2021

[BACKPORT] [FEATURE] Add API to control denormalized computations (#20387 )

committedJul 30, 2021

Commits on Jul 6, 2021

[BUGFIX] Add checks in BatchNorm's infer shape (#20415 )
bgawrych
committedJul 6, 2021