Commits

Branch selector

 main

User selector

MaximumEntropy

Datepicker

All time

Commit History

Commits on Oct 15, 2022

• Fixes some args that were not removed properly for multilingual Megatron NMT (#5142)

  Show description for a933fe6
  

  MaximumEntropy

  and

  pre-commit-ci[bot]

  authored
  a933fe6

  Copy full SHA for a933fe6

• Prompt learning of Huggingface T5v1.1 converted checkpoints (#4746)

  Show description for bc48db3
  

  9 peopleauthored
  bc48db3

  Copy full SHA for bc48db3

Commits on Oct 10, 2022

• T5 Prompt Learning Fixes for Pipeline Parallel (#5120)

  Show description for abab493
  

  4 peopleauthored
  abab493

  Copy full SHA for abab493

• Remove unsupported arguments from MegatronNMT (#5065)

  Show description for 02a36e5
  MaximumEntropy

  authored
  02a36e5

  Copy full SHA for 02a36e5

Commits on Oct 7, 2022

• Fix and refactor consumed samples save/restore for Megatron models. (#5077)

  Show description for 9782e9a
  MaximumEntropy

  authored
  9782e9a

  Copy full SHA for 9782e9a

Commits on Sep 9, 2022

• Fix ptl_load_state (#4914)

  Show description for b85fc77
  MaximumEntropy

  authored
  b85fc77

  Copy full SHA for b85fc77

Commits on Sep 1, 2022

• Fix decoding for T5 models with RPE (#4847)

  Show description for b3ed677
  MaximumEntropy

  authored
  b3ed677

  Copy full SHA for b3ed677

Commits on Aug 26, 2022

• Megatron Encoder Decoder models with RPE and PP > 2 (#4663)

  Show description for 4d3e335
  

  8 peopleauthored
  4d3e335

  Copy full SHA for 4d3e335

Commits on Aug 18, 2022

• Fixes NLPModel's load from checkpoint due to PTL private function changes (#4755)

  Show description for 8e73224
  

  MaximumEntropy

  and

  ericharper

  authored
  8e73224

  Copy full SHA for 8e73224

• Fix logging in gradient clipping with PTL 1.7.2 (#4769)

  Show description for 6abfbbf
  MaximumEntropy

  authored
  6abfbbf

  Copy full SHA for 6abfbbf

Commits on Aug 4, 2022

• Label Smoothing in VocabParallelCrossEntropy (#4602)

  Show description for b7948c1
  

  3 peopleauthored
  b7948c1

  Copy full SHA for b7948c1

Commits on Aug 2, 2022

• Add sentencepiece legacy arg to tokenizer configs (#4659)

  Show description for 5c8fe3a
  MaximumEntropy

  authored
  5c8fe3a

  Copy full SHA for 5c8fe3a

• Pin lightning (#4660)

  Show description for e1d09b0
  MaximumEntropy

  authored
  e1d09b0

  Copy full SHA for e1d09b0

• Asymmetric Encoder and Decoder Configuration for Megatron Models (#4568)

  Show description for 8e1436b
  

  MaximumEntropy

  and

  michalivne

  authored
  8e1436b

  Copy full SHA for 8e1436b

Commits on Aug 1, 2022

• T0 model and dataset (#4598)

  Show description for 1a9daa5
  

  MaximumEntropy

  and

  ericharper

  authored
  1a9daa5

  Copy full SHA for 1a9daa5

Commits on Jul 30, 2022

• Maximum sample-based training for Megatron NMT and Text Memmap based Seq2seq Pre-training (#4396)

  Show description for 0b7df7a
  

  3 peopleauthored
  0b7df7a

  Copy full SHA for 0b7df7a

Commits on Jul 28, 2022

• Fix metric setup for finetuning without a test set (#4585)

  Show description for 72d78d8
  

  MaximumEntropy

  and

  okuchaiev

  authored
  72d78d8

  Copy full SHA for 72d78d8

Commits on Jul 22, 2022

• Option to disregard document boundaries for t5, bart, ul2 (#4481)

  Show description for 6442e33
  MaximumEntropy

  authored
  6442e33

  Copy full SHA for 6442e33

Commits on Jul 15, 2022

• Add length ratio filtering script (#4551)

  Show description for 65b9b57
  

  MaximumEntropy

  and

  aklife97

  authored
  65b9b57

  Copy full SHA for 65b9b57

Commits on Jul 11, 2022

• Additional sentencepiece args - Byte fallback, split digits, split_on_whitespace (#4525)

  Show description for 7f75191
  MaximumEntropy

  authored
  7f75191

  Copy full SHA for 7f75191

Commits on Jul 6, 2022

• Megatron perceiver with tensor parallelism only (#4318)

  Show description for ab6c46b
  

  12 peopleauthored
  ab6c46b

  Copy full SHA for ab6c46b

Commits on Jun 3, 2022

• Finetune T5 on the prefix-lm objective (#4328)

  Show description for 416d033
  

  MaximumEntropy

  and

  okuchaiev

  authored
  416d033

  Copy full SHA for 416d033

Commits on Jun 2, 2022

• Removes debug logging statements in Megatron NMT (#4312)

  Show description for c1ae0a7
  MaximumEntropy

  authored
  c1ae0a7

  Copy full SHA for c1ae0a7

Commits on Jun 1, 2022

• Binarized memmap dataloader for Megatron NMT, Inference and checkpoint -> nemo (#4137)

  Show description for 1f4c744
  

  3 peopleauthored
  1f4c744

  Copy full SHA for 1f4c744

• Configure T5 finetuning metrics (#4122)

  Show description for f9d45db
  MaximumEntropy

  authored
  f9d45db

  Copy full SHA for f9d45db

• Implements the UL2 Dataset and config (#4184)

  Show description for 2af3786
  MaximumEntropy

  authored
  2af3786

  Copy full SHA for 2af3786

Commits on May 27, 2022

• Fixes for Megatron Base Model Artifacts (#4248)

  Show description for 05f9ba3
  MaximumEntropy

  authored
  05f9ba3

  Copy full SHA for 05f9ba3

Commits on May 26, 2022

• BigNLP perf regression fix (#4267)

  Show description for 8329910
  MaximumEntropy

  authored
  8329910

  Copy full SHA for 8329910

Commits on May 4, 2022

• Megatron NMT Restore from T5/BART and finetune (#3977)

  Show description for dc13c5d
  

  MaximumEntropy

  and

  aklife97

  authored
  dc13c5d

  Copy full SHA for dc13c5d

Commits on Apr 29, 2022

• T5 finetuning for generic small text-to-text datasets (#4032)

  Show description for 629e995
  

  4 peopleauthored
  629e995

  Copy full SHA for 629e995

• RMSNorm, Normformer and fixes from merging 1.8.0 into main (#4048)

  Show description for 4c14b20
  

  MaximumEntropy

  and

  ericharper

  authored
  4c14b20

  Copy full SHA for 4c14b20

Commits on Apr 19, 2022

• Exp manager log rank 0 only arguments (#4026)

  Show description for 48b2b05
  MaximumEntropy

  authored
  48b2b05

  Copy full SHA for 48b2b05

Commits on Apr 15, 2022

• Add NMT method to translate with TN/ITN pre/post-processing (#4009)

  Show description for a1c2d4e
  

  MaximumEntropy

  and

  aklife97

  authored
  a1c2d4e

  Copy full SHA for a1c2d4e

• Option to remove bias terms from Megatron transformers (#3973)

  Show description for c978f4f
  MaximumEntropy

  authored
  c978f4f

  Copy full SHA for c978f4f

Commits on Apr 7, 2022

• Glu activation variants (#3951)

  Show description for afc7b71
  MaximumEntropy

  authored
  afc7b71

  Copy full SHA for afc7b71

Pagination

Previous

Next