Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on Mar 2, 2020

• Revert "remove gencode from multihead_attn build (NVIDIA#731)"

  Show description for 5633f6d
  ptrblck

  committed
  5633f6d

  Copy full SHA for 5633f6d

Commits on Feb 27, 2020

• NHWC support for multi tensor apply (NVIDIA#732)

  Show description for de6378f
  mcarilli

  authored
  de6378f

  Copy full SHA for de6378f

Commits on Feb 25, 2020

• remove gencode from multihead_attn build (NVIDIA#731)

  ptrblck

  authored
  92b3b9a

  Copy full SHA for 92b3b9a

• remove duplicated multihead_attn install (NVIDIA#729)

  ptrblck

  authored
  5f6b9b0

  Copy full SHA for 5f6b9b0

• Adding 'ctc_loss' to the list of FP32 funcs (NVIDIA#722)

  saranshkarira

  authored
  93cabd5

  Copy full SHA for 93cabd5

Commits on Feb 24, 2020

• Change to Multihead Attention to allow Batched GEMMs larger than 64K. (NVIDIA#728)

  Show description for 1733946
  kevinstephano

  authored
  1733946

  Copy full SHA for 1733946

Commits on Feb 15, 2020

• change include_dirs to abs path (NVIDIA#719)

  FDecaYed

  authored
  50338df

  Copy full SHA for 50338df

Commits on Feb 10, 2020

• Fix opt_level command line arg in instructions. (NVIDIA#713)

  Show description for 5b71d36
  a-y-khan

  authored
  5b71d36

  Copy full SHA for 5b71d36

Commits on Feb 6, 2020

• Add Fast Multihead Attention to APEX Contrib (NVIDIA#697)

  Show description for 3f94528
  kevinstephano

  authored
  3f94528

  Copy full SHA for 3f94528

Commits on Feb 5, 2020

• Fix attribute name mismatch in state_dict() and load_state_dict() (NVIDIA#704)

  Show description for 494f8ab
  kexinyu

  authored
  494f8ab

  Copy full SHA for 494f8ab

Commits on Jan 27, 2020

• Channels last support (NVIDIA#668)

  

  VitalyFedyunin

  authored and

  mcarilli

  committed
  2ca894d

  Copy full SHA for 2ca894d

Commits on Jan 21, 2020

• removing build target sm_70 from bnp (NVIDIA#683)

  

  jjsjann123

  authored and

  ptrblck

  committed
  b66ffc1

  Copy full SHA for b66ffc1

Commits on Jan 8, 2020

• add WAR for pip>=19.3.1 (NVIDIA#652)

  Show description for b5a7c5f
  

  ptrblck

  authored and

  mcarilli

  committed
  b5a7c5f

  Copy full SHA for b5a7c5f

Commits on Dec 18, 2019

• fix beta2 => beta1 (NVIDIA#661)

  

  bonlime

  authored and

  mcarilli

  committed
  0ce8ad3

  Copy full SHA for 0ce8ad3

• updated apex.contrib.optimizers.FP16_Optimizer and FusedSGD (NVIDIA#657)

  

  kexinyu

  authored and

  mcarilli

  committed
  c19ee27

  Copy full SHA for c19ee27

Commits on Dec 5, 2019

• Fixing typo in PyProf README (NVIDIA#637)

  

  ntenenz

  authored and

  mcarilli

  committed
  4ad9b3b

  Copy full SHA for 4ad9b3b

Commits on Dec 3, 2019

• Don't check if distributed is initialized on Windows

  definitelynotmcarilli

  committed
  f37fdf0

  Copy full SHA for f37fdf0

Commits on Nov 22, 2019

• update _amp_state to check distributed on maybe_print (NVIDIA#620)

  

  rmrao

  authored and

  mcarilli

  committed
  82dac9c

  Copy full SHA for 82dac9c

Commits on Nov 6, 2019

• fixing batchnorm 1d input (NVIDIA#590)

  

  jjsjann123

  authored and

  mcarilli

  committed
  37cdaf4

  Copy full SHA for 37cdaf4

Commits on Oct 30, 2019

• Update README.md

  mcarilli

  authored
  606c3dc

  Copy full SHA for 606c3dc

Commits on Oct 23, 2019

• add gelu activation to fp32 list (NVIDIA#564)

  

  Bram Vanroy

  authored and

  mcarilli

  committed
  5b29cc1

  Copy full SHA for 5b29cc1

Commits on Oct 22, 2019

• Making the encouragement to use O1 a bit stronger...

  definitelynotmcarilli

  committed
  95d6c00

  Copy full SHA for 95d6c00

Commits on Oct 19, 2019

• Made the patched optimizer step function a full method, not simply a function stored as an instance member (NVIDIA#553)

  

  aronhoff

  authored and

  mcarilli

  committed
  4b91326

  Copy full SHA for 4b91326

Commits on Oct 10, 2019

• Adding presentation link to sphinx landing page

  definitelynotmcarilli

  committed
  0889859

  Copy full SHA for 0889859

• Adding links to references. (maybe make this a subrepo?)

  mcarilli

  authored
  ec93c75

  Copy full SHA for ec93c75

Commits on Oct 9, 2019

• allow for non-distributed envs (Windows) (NVIDIA#531)

  

  Bram Vanroy

  authored and

  mcarilli

  committed
  fab319f

  Copy full SHA for fab319f

• Fixed tensor core lookup for Turing (NVIDIA#534)

  

  mkolod

  authored and

  mcarilli

  committed
  753c427

  Copy full SHA for 753c427

Commits on Oct 8, 2019

• Include loss scaling in README code example (NVIDIA#523)

  

  f0k

  authored and

  mcarilli

  committed
  e87b579

  Copy full SHA for e87b579

Commits on Oct 4, 2019

• move previous fused_adam and fp16_optimizer to contrib (NVIDIA#517)

  Show description for 1904e48
  

  FDecaYed

  authored and

  mcarilli

  committed
  1904e48

  Copy full SHA for 1904e48

Commits on Oct 3, 2019

• Disable tests for mixed opt_levels, add bitwise accurate test of parameters (NVIDIA#520)

  Show description for 0b74bfd
  

  ptrblck

  authored and

  mcarilli

  committed
  0b74bfd

  Copy full SHA for 0b74bfd

Commits on Oct 2, 2019

• fix facebookresearch/maskrcnn-benchmark#802 (NVIDIA#516)

  

  timotheecour

  authored and

  mcarilli

  committed
  03421e8

  Copy full SHA for 03421e8

Commits on Sep 13, 2019

• Seems to work locally (NVIDIA#490)

  mcarilli

  authored
  3ae89c7

  Copy full SHA for 3ae89c7

Commits on Sep 12, 2019

• Fixed error in convert_syncbn_model function (NVIDIA#380)

  

  whyjay

  authored and

  mcarilli

  committed
  e6cb749

  Copy full SHA for e6cb749

Commits on Sep 11, 2019

• removing nvtx range used for debugging (NVIDIA#485)

  

  jjsjann123

  authored and

  mcarilli

  committed
  ad98cc5

  Copy full SHA for ad98cc5

Commits on Sep 6, 2019

• Fix for NVIDIA#456 (NVIDIA#477)

  Show description for 325f5a0
  mcarilli

  authored
  325f5a0

  Copy full SHA for 325f5a0

Pagination

Previous

Next