Commits

Breadcrumbs

History for

on32c67c2

User selector

All users

Datepicker

All time

Commit History

Commits on Dec 18, 2020

• Deprecating Horovod and refactored Adasum computations (microsoft#5468)

  Show description for 32c67c2
  

  Tixxx

  authored
  32c67c2

  Copy full SHA for 32c67c2

Commits on Dec 17, 2020

• MLAS: add NEON version of int8 depthwise convolution (microsoft#6152)

  tracysh

  authored
  503b61d

  Copy full SHA for 503b61d

Commits on Dec 16, 2020

• Fix edge case in BFCArena where allocation failures could lead to an infinite loop. (microsoft#6145)

  Show description for 7250562
  skottmckay

  authored
  7250562

  Copy full SHA for 7250562

• Fix TensorRT kernel conflict issue for subgraphs of control flow operators (microsoft#6115)

  Show description for aa49e47
  stevenlix

  authored
  aa49e47

  Copy full SHA for aa49e47

• Fix CUDA test hang: (microsoft#6138)

  Show description for 0978d2b
  toothache

  authored
  0978d2b

  Copy full SHA for 0978d2b

• nnapi add min max support (microsoft#6117)

  guoyu-wang

  authored
  b648bf6

  Copy full SHA for b648bf6

• Unify IExecutionProvider and IExecutionProviderFactory interfaces (microsoft#6108)

  Show description for ac62cf8
  RyanUnderhill

  authored
  ac62cf8

  Copy full SHA for ac62cf8

Commits on Dec 15, 2020

• Model Fusion For Bart (microsoft#6105)

  Show description for 980a93c
  liuziyue

  authored
  980a93c

  Copy full SHA for 980a93c

• Deprecate Python global configuration functions [Part 1] (microsoft#5923)

  Show description for 64709b1
  edgchen1

  authored
  64709b1

  Copy full SHA for 64709b1

• Minor changes to AMD element-wise kernels to converge with CUDA element-wise kernels.

  jessebenson

  committed
  a8d549e

  Copy full SHA for a8d549e

Commits on Dec 14, 2020

• Reduce amount of compiled CUDA device code (microsoft#6118)

  Show description for 9810b9e
  edgchen1

  authored
  9810b9e

  Copy full SHA for 9810b9e

• Unify activation and initializer alignment value (microsoft#6109)

  Show description for eb5c1f0
  SherlockNoMad

  authored
  eb5c1f0

  Copy full SHA for eb5c1f0

Commits on Dec 12, 2020

• Honor auto_pad attribute in ConvTranspose (microsoft#4271)

  hariharans29

  authored
  c755ca0

  Copy full SHA for c755ca0

Commits on Dec 11, 2020

• Fix multi-tensor LAMB reduction to be deterministic (microsoft#6028)

  Show description for 6cb5d3a
  

  Suffian Khan

  authored
  6cb5d3a

  Copy full SHA for 6cb5d3a

• Fix DEBUG_NODE_INPUTS_OUTPUTS test by putting it in a separate process, clean up unused test_main.cc files. (microsoft#5949)

  Show description for c8ac34d
  edgchen1

  authored
  c8ac34d

  Copy full SHA for c8ac34d

• Introduce VariadicAlias, remove hardcoded alias limits (microsoft#6106)

  Show description for a53f4dd
  

  SherlockNoMad

  and

  Sherlock Huang

  authored
  a53f4dd

  Copy full SHA for a53f4dd

• Make ROCM and CUDA reduction_all code more similar.

  jessebenson

  committed
  38c49c2

  Copy full SHA for 38c49c2

Commits on Dec 10, 2020

• adding fp16 support for topk cuda kernel (microsoft#6082)

  Show description for e945b5f
  

  duli2012

  and

  Du Li

  authored
  e945b5f

  Copy full SHA for e945b5f

Commits on Dec 9, 2020

• Enable varied input type for custom op (microsoft#6066)

  Show description for 404982d
  RandySheriffH

  authored
  404982d

  Copy full SHA for 404982d

• Update AMD transpose to match CUDA transpose.

  jessebenson

  committed
  cc47cfc

  Copy full SHA for cc47cfc

• Reduction kernel optimization (microsoft#6088)

  Show description for abdbb5f
  edgchen1

  authored
  abdbb5f

  Copy full SHA for abdbb5f

• clean un-used code. (microsoft#6059)

  Show description for d95fc5e
  weixingzhang

  authored
  d95fc5e

  Copy full SHA for d95fc5e

• Enable passing initial optimizer state while creating training session (microsoft#5869)

  Show description for b1a75d0
  ashbhandare

  authored
  b1a75d0

  Copy full SHA for b1a75d0

Commits on Dec 8, 2020

• Update version to 1.6.0 (microsoft#6041)

  Show description for 3e81711
  

  duli2012

  and

  Du Li

  authored
  3e81711

  Copy full SHA for 3e81711

• Bump up API version for 1.6 release (microsoft#6076)

  pranavsharma

  authored
  2c5ba9a

  Copy full SHA for 2c5ba9a

• Support export >2G model when using optimizer.py only (microsoft#6014)

  Show description for fa06be2
  wangyems

  authored
  fa06be2

  Copy full SHA for fa06be2

Commits on Dec 7, 2020

• Update profiler tool to support gpt2 and longformer models (microsoft#6011)

  Show description for 51fbe87
  tianleiwu

  authored
  51fbe87

  Copy full SHA for 51fbe87

• Remove python 3.8 Windows GPU build from python packaging pipeline (microsoft#6054)

  Show description for 925879a
  snnn

  authored
  925879a

  Copy full SHA for 925879a

• fix windows cuda support for python 3.8 + (microsoft#6046)

  Show description for 020efc9
  jywu-msft

  authored
  020efc9

  Copy full SHA for 020efc9

• Improve checkpointing for Zero stage 1 (microsoft#5478)

  Show description for 7cebf76
  ashbhandare

  authored
  7cebf76

  Copy full SHA for 7cebf76

Commits on Dec 3, 2020

• Use __launch_bounds__ workaround, rather than limiting threads to 256 on AMD.

  jessebenson

  committed
  14f6eb1

  Copy full SHA for 14f6eb1

• Fix AMD multi-tensor implementation.

  jessebenson

  committed
  245d436

  Copy full SHA for 245d436

• longformer onnx conversion and benchmark tools (microsoft#6007)

  Show description for cdb9120
  tianleiwu

  authored
  cdb9120

  Copy full SHA for cdb9120

• Support Fusion for 1 and 2 Inputs Bert Models Converted From tf (microsoft#5993)

  Show description for 3b198c9
  liuziyue

  authored
  3b198c9

  Copy full SHA for 3b198c9

• Fix build.py bug which prevents running some unit tests (microsoft#5990)

  Show description for 2b35f7d
  baijumeswani

  authored
  2b35f7d

  Copy full SHA for 2b35f7d

Pagination

Previous

Next