Commits

Breadcrumbs

History for

onmaster

User selector

All users

Datepicker

All time

Commit History

Commits on Jun 2, 2023

• [Test][Transformer] Pre-parse container version (NVIDIA#1673)

  Show description for 05091d4
  Aidyn-A

  authored
  05091d4

Commits on May 12, 2023

• allow for custom directory of xml reports (NVIDIA#1661)

  Show description for 2d876b8
  crcrpar

  authored
  2d876b8

Commits on Apr 27, 2023

• Using regular layernorm for ONNX and TS export (NVIDIA#1634)

  Show description for 85e9edd
  

  borisfom

  and

  crcrpar

  authored
  85e9edd

Commits on Apr 26, 2023

• [Transformer][test/L0] Increase max world_size for BERT and GPT (NVIDIA#1651)

  Aidyn-A

  authored
  5b5d410

Commits on Apr 14, 2023

• Add support for FP32 master weights in FusedAdam optimizer (NVIDIA#1625)

  Show description for 6bd01c4
  

  5 peopleauthored
  6bd01c4

Commits on Mar 23, 2023

• check for empty params in group (NVIDIA#1596)

  Show description for 0420735
  

  3 peopleauthored
  0420735

Commits on Mar 22, 2023

• Revert "Add support for FP32 master weights in FusedAdam optimizer (NVIDIA#1623)" (NVIDIA#1624)

  Show description for 4060aef
  crcrpar

  authored
  4060aef

• Add support for FP32 master weights in FusedAdam optimizer (NVIDIA#1623)

  Show description for 6952004
  

  4 peopleauthored
  6952004

Commits on Jan 31, 2023

• Capturable Adam optimizer (NVIDIA#1560)

  Show description for ad68a2f
  gilshm

  authored
  ad68a2f

Commits on Jan 26, 2023

• update test routines (NVIDIA#1567)

  Aidyn-A

  authored
  537424d

Commits on Dec 8, 2022

• Refactor run_transformer bert minimal and gpt minimal tests (NVIDIA#1540)

  Show description for 85af77c
  Fuzzkatt

  authored
  85af77c

Commits on Nov 22, 2022

• Update megatron fused softmax follow megatron-lm (NVIDIA#1539)

  Show description for abeca58
  

  yaoyu-33

  and

  Yu Yao

  authored
  abeca58

Commits on Oct 27, 2022

• update exception message (NVIDIA#1524)

  Show description for 6a40a0a
  crcrpar

  authored
  6a40a0a

• update run_transformer tests to default to using pytorch native UCC instead of torch_ucc (NVIDIA#1495)

  Show description for 0d06a73
  Fuzzkatt

  authored
  0d06a73

• Support overlapped grad sync with Megatron interleaved pipeline parallelism (NVIDIA#1514)

  Show description for f683961
  timmoon10

  authored
  f683961

Commits on Oct 13, 2022

• add tearDown (NVIDIA#1508)

  Aidyn-A

  authored
  2c3b8f9

Commits on Oct 12, 2022

• [transformer] Allow for skipping stream sync (NVIDIA#1505)

  Show description for 806f9b0
  

  3 peopleauthored
  806f9b0

Commits on Oct 10, 2022

• Deprecate outdated features (NVIDIA#1506)

  Show description for 48bf0df
  crcrpar

  authored
  48bf0df

Commits on Sep 20, 2022

• [contrib][DistributedFusedAdam] Support overlapped grad sync with Megatron pipeline parallelism (NVIDIA#1475)

  Show description for 2b0e837
  timmoon10

  authored
  2b0e837

Commits on Sep 16, 2022

• introducing APEX_RUN_WITH_SLOW_TESTS env var (NVIDIA#1489)

  Show description for 6b5405e
  crcrpar

  authored
  6b5405e

Commits on Sep 14, 2022

• Update test_fused_layer_norm.py (NVIDIA#1485)

  Show description for 136a13c
  Fuzzkatt

  authored
  136a13c

Commits on Sep 9, 2022

• Update apex.mlp to use fp16 in autocast (NVIDIA#1477)

  Show description for 81c9aca
  crcrpar

  authored
  81c9aca

Commits on Aug 24, 2022

• Label smoothing in vocab parallel cross entropy (NVIDIA#1457)

  Show description for 8cea0b4
  MaximumEntropy

  authored
  8cea0b4

Commits on Aug 15, 2022

• Use xmlrunner.XMLTestRunner accordingly in tests/L0/run_test.py (NVIDIA#1451)

  Show description for e75cb25
  crcrpar

  authored
  e75cb25

Commits on Aug 9, 2022

• Update mlp_cuda test (NVIDIA#1425)

  Show description for 71e5871
  crcrpar

  authored
  71e5871

• [transformer] unittest: less mem consumption for generic softmax (NVIDIA#1448)

  Show description for 2f67d22
  crcrpar

  authored
  2f67d22

Commits on Aug 5, 2022

• Add a generic fused softmax (NVIDIA#1440)

  Show description for cd0a1f1
  yidong72

  authored
  cd0a1f1

Commits on Jul 26, 2022

• Fix bug when initializing model-parallel process groups for GPT-3 (NVIDIA#1435)

  Show description for fb21698
  timmoon10

  authored
  fb21698

Commits on Jul 25, 2022

• [transformer] update tests (NVIDIA#1428)

  Aidyn-A

  authored
  e57d9e7

Commits on Jul 20, 2022

• [transformer] UCC async test (NVIDIA#1417)

  Show description for a29a698
  Aidyn-A

  authored
  a29a698

Commits on Jul 14, 2022

• Time dimension shape check for fused scale mask softmax kernel (NVIDIA#1421)

  Show description for 1337e81
  MaximumEntropy

  authored
  1337e81

Commits on Jul 11, 2022

• update: mpu for t5 rpe (NVIDIA#1416)

  Show description for 5ff5a88
  

  PerkzZheng

  and

  crcrpar

  authored
  5ff5a88

Commits on Jul 7, 2022

• Remove pyprof and reparameterization (NVIDIA#1404)

  Show description for 8a7a332
  crcrpar

  authored
  8a7a332

Commits on Jun 23, 2022

• [transformer] Port Sequence Parallelism (takeover of NVIDIA#1396) (NVIDIA#1400)

  Show description for 3ff1a10
  crcrpar

  authored
  3ff1a10

• Move distributed Adam unit test to contrib dir (NVIDIA#1406)

  Show description for 57f890a
  timmoon10

  authored
  57f890a

Pagination

PreviousNext