Commits

Breadcrumbs

History for

onmain

User selector

All users

Datepicker

All time

Commit History

Commits on May 25, 2024

• Change default to use mcore models, not legacy.

  jaredcasper

  committed
  9ad1a56

  Copy full SHA for 9ad1a56

Commits on May 2, 2024

• Merge branch 'boxiangw/mlperf-option-add-one-extra-token' into 'main'

  Show description for db3a3f7
  ericharper

  committed
  db3a3f7

  Copy full SHA for db3a3f7

• [MLPerf] GPT dataset features: drop last partial validation sequence, drop extra token, return sample with 1s loss mask, mock dataset testing

  

  BoxiangW

  authored and

  ericharper

  committed
  c90aa16

  Copy full SHA for c90aa16

Commits on Apr 29, 2024

• Fix Cross Entropy Loss Averaging

  

  Mike Chrzanowski

  authored and

  jaredcasper

  committed
  adfa873

  Copy full SHA for adfa873

Commits on Apr 17, 2024

• Support missing size and missing weights in dataset creation and blending

  

  John Kamalu

  authored and

  ericharper

  committed
  dc52e84

  Copy full SHA for dc52e84

Commits on Mar 26, 2024

• Merge branch 'jcasper/refactor' into 'main'

  Show description for da0afe0
  jaredcasper

  committed
  da0afe0

  Copy full SHA for da0afe0

• Refactor everything outside of core to be out of the main megatron. namespace.

  jaredcasper

  committed
  38644dd

  Copy full SHA for 38644dd

• move is_built_on_rank from config to builder

  

  John Kamalu

  authored and

  ericharper

  committed
  9475bab

  Copy full SHA for 9475bab

Commits on Jan 30, 2024

• Merge branch 'main' into 'local_spec_bert'

  Show description for 329baac
  shanmugamr1992

  committed
  329baac

  Copy full SHA for 329baac

• Add support for masked WordPiece datasets BERT and T5

  

  John Kamalu

  authored and

  ericharper

  committed
  25a9946

  Copy full SHA for 25a9946

Commits on Jan 26, 2024

• Adding bert local spec test

  

  Shanmugam Ramasamy

  committed
  fc316ff

  Copy full SHA for fc316ff

Commits on Jan 25, 2024

• Adding bert local spec test

  

  Shanmugam Ramasamy

  committed
  e6ef9ea

  Copy full SHA for e6ef9ea

• Adding bert local spec test

  

  Shanmugam Ramasamy

  committed
  e25970f

  Copy full SHA for e25970f

Commits on Nov 16, 2023

• Retro for Megatron Core

  

  lmcafee-nvidia

  authored and

  jaredcasper

  committed
  67a0e5d

  Copy full SHA for 67a0e5d

Commits on Oct 30, 2023

• Addressed jared's comments

  

  Shanmugam Ramasamy

  committed
  fa56f2b

  Copy full SHA for fa56f2b

Commits on Oct 26, 2023

• Refactor dataset code and move to core

  

  John Kamalu

  authored and

  jaredcasper

  committed
  a0e190c

  Copy full SHA for a0e190c

Commits on Oct 22, 2023

• Addressing jared's comments

  shanmugamr1992

  committed
  da169da

  Copy full SHA for da169da

• Addressing jared's comments

  shanmugamr1992

  committed
  4ef4555

  Copy full SHA for 4ef4555

• Addressing jared's comments

  shanmugamr1992

  committed
  12f4669

  Copy full SHA for 12f4669

Commits on Oct 19, 2023

• Refactoring bert

  shanmugamr1992

  committed
  2f3afc7

  Copy full SHA for 2f3afc7

Commits on Oct 18, 2023

• Refactoring bert

  shanmugamr1992

  committed
  40c2b52

  Copy full SHA for 40c2b52

• Refactoring bert

  shanmugamr1992

  committed
  2f7c839

  Copy full SHA for 2f7c839

Commits on Sep 26, 2023

• Adding refactored bert

  

  root

  committed
  0b4e362

  Copy full SHA for 0b4e362

• Refactoring bert for recent changes

  

  Shanmugam Ramasamy

  committed
  273f086

  Copy full SHA for 273f086

Commits on Sep 7, 2023

• Document, clean, and refactor megatron/data for GPT

  

  John Kamalu

  authored and

  jaredcasper

  committed
  34ac1ad

  Copy full SHA for 34ac1ad

Commits on Jul 11, 2023

• Merge branch 'main' into zhuoliny/flamingo-dev

  jaredcasper

  committed
  e5fdfd7

  Copy full SHA for e5fdfd7

Commits on Jun 25, 2023

• code refactor + packing text and img in to a single bin file

  kvareddy

  committed
  efd8f78

  Copy full SHA for efd8f78

Commits on May 31, 2023

• Convert bert to use config, some cleanup of module.

  jaredcasper

  committed
  02fffd2

  Copy full SHA for 02fffd2

Commits on Mar 23, 2023

• Move pipeline parallel functionality into core with associated changes.

  jaredcasper

  committed
  3c92fa9

  Copy full SHA for 3c92fa9

Commits on Sep 24, 2022

• Alias core.parallel_state as mpu and use it throughout code. RIP mpu.

  jaredcasper

  committed
  5942af9

  Copy full SHA for 5942af9

Commits on Sep 23, 2022

• Clean up licensing.

  Show description for fabad46
  jaredcasper

  committed
  fabad46

  Copy full SHA for fabad46

Commits on Aug 10, 2022

• Timing levels

  

  shoeybi

  authored and

  jaredcasper

  committed
  77efccc

  Copy full SHA for 77efccc

Commits on Jul 30, 2021

• Support for pipeline parallelism in T5 model

  Show description for 46c74b4
  deepakn94

  committed
  46c74b4

  Copy full SHA for 46c74b4

Commits on Apr 16, 2021

• Integrate code from t5_main into existing code.

  jaredcasper

  committed
  48a5e0d

  Copy full SHA for 48a5e0d

Commits on Mar 24, 2021

• pipeline code simplification

  

  kvareddy

  authored and

  jaredcasper

  committed
  3b91262

  Copy full SHA for 3b91262

Pagination

Previous

Next