Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on Mar 24, 2021

• sync-free Distributed LAMB + parameter reordering (NVIDIA#1055)

  Show description for a651e2c
  

  kexinyu

  and

  Kexin Yu

  authored
  a651e2c

  Copy full SHA for a651e2c

• Initial check-in of the transducer extensions (NVIDIA#1069)

  Show description for d86d1b0
  nanz-nv

  authored
  d86d1b0

  Copy full SHA for d86d1b0

Commits on Feb 23, 2021

• fast layer norm (NVIDIA#1037)

  yjk21

  authored
  e2083df

  Copy full SHA for e2083df

Commits on Feb 10, 2021

• fix import container_abcs issue (NVIDIA#1049)

  Show description for a78ccf0
  ShoufaChen

  authored
  a78ccf0

  Copy full SHA for a78ccf0

Commits on Jan 20, 2021

• cuda rng changes for graph capture with apex MHA (NVIDIA#1025)

  Show description for eefb1ba
  

  seryilmaz

  and

  Sukru Eryilmaz

  authored
  eefb1ba

  Copy full SHA for eefb1ba

Commits on Dec 17, 2020

• Merge pull request NVIDIA#1015 from jpool-nv/patch-1

  Show description for 154c633
  thorjohnsen

  authored
  154c633

  Copy full SHA for 154c633

• Update ASP README to highlight default recipe

  Show description for 56914d4
  jpool-nv

  authored
  56914d4

  Copy full SHA for 56914d4

Commits on Dec 4, 2020

• remove noise pip-version-check noise that hides the outcome of the build (NVIDIA#998)

  stas00

  authored
  8cf5ae6

  Copy full SHA for 8cf5ae6

• Distributed LAMB fixes (NVIDIA#1007)

  Show description for 8a80d47
  

  3 peopleauthored
  8a80d47

  Copy full SHA for 8a80d47

• Seryilmaz/fused dropout softmax (NVIDIA#985)

  Show description for 3fe10b5
  seryilmaz

  authored
  3fe10b5

  Copy full SHA for 3fe10b5

Commits on Dec 2, 2020

• Fix lack of proper loading of best_prec1 from the checkpoint (NVIDIA#1000)

  Show description for 6c186b3
  JanuszL

  authored
  6c186b3

  Copy full SHA for 6c186b3

Commits on Dec 1, 2020

• DistributedFusedAdam Model Parallelism Support (Megatron) (NVIDIA#981)

  Show description for 6b7e77b
  

  3 peopleauthored
  6b7e77b

  Copy full SHA for 6b7e77b

Commits on Oct 19, 2020

• Optimize the sync batchnorm by batching the communication (NVIDIA#980)

  Show description for 8a1ed9e
  lly-zero-one

  authored
  8a1ed9e

  Copy full SHA for 8a1ed9e

Commits on Sep 29, 2020

• use reshape instead of view (NVIDIA#971)

  ptrblck

  authored
  a109f85

  Copy full SHA for a109f85

Commits on Sep 15, 2020

• Merge pull request NVIDIA#959 from a-maci/update-ASP-readme

  Show description for 4a1fa2c
  thorjohnsen

  authored
  4a1fa2c

  Copy full SHA for 4a1fa2c

Commits on Sep 14, 2020

• Merge pull request NVIDIA#5 from a-maci/a-maci-patch-update-asp-readme

  Show description for 48fc613
  a-maci

  authored
  48fc613

  Copy full SHA for 48fc613

• Update README for ASP

  Show description for e3794f4
  a-maci

  authored
  e3794f4

  Copy full SHA for e3794f4

Commits on Aug 15, 2020

• Should pass stricter stride/size checks in pytorch (NVIDIA#942)

  mcarilli

  authored
  4ef930c

  Copy full SHA for 4ef930c

Commits on Aug 10, 2020

• move sm80 code inside MHA (NVIDIA#937)

  Show description for 5d9b5cb
  ptrblck

  authored
  5d9b5cb

  Copy full SHA for 5d9b5cb

Commits on Aug 6, 2020

• sgd supports zero-grad (NVIDIA#926)

  

  ngimel

  authored
  85b1783

  Copy full SHA for 85b1783

Commits on Aug 5, 2020

• set device guard for multi tensor optimizer implementations (NVIDIA#927)

  Show description for 274cc06
  

  ngimel

  authored
  274cc06

  Copy full SHA for 274cc06

Commits on Aug 1, 2020

• Add sm80 for CUDA >= 11 (NVIDIA#925)

  ptrblck

  authored
  5b53121

  Copy full SHA for 5b53121

Commits on Jul 30, 2020

• fix default mode missing additive mask option (NVIDIA#924)

  Show description for 700d682
  

  seryilmaz

  and

  Sukru Eryilmaz

  authored
  700d682

  Copy full SHA for 700d682

Commits on Jul 23, 2020

• Merge pull request NVIDIA#918 from a-maci/ASP_sparse_param_dict_update

  Show description for 459de22
  thorjohnsen

  authored
  459de22

  Copy full SHA for 459de22

Commits on Jul 22, 2020

• Merge pull request NVIDIA#4 from a-maci/ASP_sparse_param_dict_update

  Show description for eb5e96c
  a-maci

  authored
  eb5e96c

  Copy full SHA for eb5e96c

• Accept custom (layer type:param name) to include in sparse_parameter dictionary

  Show description for b3c1641
  a-maci

  authored
  b3c1641

  Copy full SHA for b3c1641

• Merge pull request NVIDIA#3 from NVIDIA/master

  Show description for eb95950
  a-maci

  authored
  eb95950

  Copy full SHA for eb95950

Commits on Jul 21, 2020

• Merge pull request NVIDIA#917 from a-maci/master

  Show description for 0ac5dd6
  thorjohnsen

  authored
  0ac5dd6

  Copy full SHA for 0ac5dd6

Commits on Jul 20, 2020

• Merge pull request NVIDIA#2 from a-maci/a-maci-patch-1

  Show description for 089149d
  a-maci

  authored
  089149d

  Copy full SHA for 089149d

• Fixing mask multiplication with grad tensors

  Show description for 774de91
  a-maci

  authored
  774de91

  Copy full SHA for 774de91

• Merge pull request NVIDIA#1 from NVIDIA/master

  Show description for 3dd3607
  a-maci

  authored
  3dd3607

  Copy full SHA for 3dd3607

Commits on Jul 16, 2020

• Merge pull request NVIDIA#910 from szmigacz/smigacz/mha_xavier_init_gain_fix

  Show description for 3104fd5
  thorjohnsen

  authored
  3104fd5

  Copy full SHA for 3104fd5

• Merge pull request NVIDIA#904 from ksivaman/sparsity

  Show description for 4027bcb
  thorjohnsen

  authored
  4027bcb

  Copy full SHA for 4027bcb

Commits on Jul 9, 2020

• Fixed weight init for fused weight matrices in fused MHA by adding correct gain factor.

  szmigacz

  committed
  a0d99fd

  Copy full SHA for a0d99fd

Commits on Jul 6, 2020

• [sync BN] (NVIDIA#792)

  Show description for 1ff54b8
  jjsjann123

  authored
  1ff54b8

  Copy full SHA for 1ff54b8

Pagination

Previous

Next