Commits

Breadcrumbs

History for

onmain

User selector

All users

Datepicker

All time

Commit History

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

Commits on Feb 9, 2021

• Interleaved pipeline execution and code refactoring

  Show description for dd88906
  deepakn94

  committed
  dd88906

  Copy full SHA for dd88906

Commits on Jan 25, 2021

• Adding option to remove the binary head for BERT

  

  shoeybi

  authored and

  jaredcasper

  committed
  dcff1ac

  Copy full SHA for dcff1ac

Commits on Dec 19, 2020

• Fix TensorBoard writes

  

  shoeybi

  authored and

  deepakn94

  committed
  b81cad6

  Copy full SHA for b81cad6

Commits on Nov 12, 2020

• Move division of loss tensor by number of microbatches to training.py

  deepakn94

  committed
  c671de3

  Copy full SHA for c671de3

• Refactor code according to Jared's comments: move pipelining and non-pipelining training loops into separate methods

  Show description for 1979c24
  deepakn94

  committed
  1979c24

  Copy full SHA for 1979c24

• Divide gradient by number of microbatches in minibatch

  deepakn94

  committed
  3d7194c

  Copy full SHA for 3d7194c

• Intra-layer MP -> Tensor MP, Inter-layer MP -> Pipeline MP

  deepakn94

  committed
  52a5f2f

  Copy full SHA for 52a5f2f

• Pipeline parallelism implementation with periodic full-pipeline syncs

  Show description for 7abd3e9
  deepakn94

  committed
  7abd3e9

  Copy full SHA for 7abd3e9

Commits on Jul 7, 2020

• Addressed Jared's comments

  kantneel

  committed
  8d7f508

  Copy full SHA for 8d7f508

Commits on Jun 26, 2020

• Merge branch 'master' into ict-merge

  kantneel

  committed
  ac967fa

  Copy full SHA for ac967fa

Commits on Jun 24, 2020

• More ict_merge changes and interactive testing

  kantneel

  committed
  3354081

  Copy full SHA for 3354081

Commits on Jun 5, 2020

• addressed jareds comments

  shoeybi

  committed
  197c132

  Copy full SHA for 197c132

• added fp16 lm cross entropy to bert

  shoeybi

  committed
  7802200

  Copy full SHA for 7802200

Pagination

Previous

Next