Code

T5 encoder decoder#249

Code

Open

ItzVladick wants to merge 122 commits intomasterallenai/longformer:masterfrom t5-encoder-decoderallenai/longformer:t5-encoder-decoderCopy head branch name to clipboard

+1,990-79

Commits

Commits on Jul 4, 2020

• adding output_attentions arg

  ibeltagy

  committed
  1f55379

• adding gradient_checkpointing config

  ibeltagy

  committed
  b98d191

• convert bart to longformer_encoder_decoder + memory profiler

  ibeltagy

  committed
  c10277c

• reqs and init

  ibeltagy

  committed
  e29d7f5

• fix req

  ibeltagy

  committed
  dd0dc0d

• req

  ibeltagy

  committed
  54a1328

Commits on Jul 6, 2020

• Update README.md

  ibeltagy

  authored
  243cfe8

• Update README.md

  ibeltagy

  authored
  fbbc770

Commits on Jul 16, 2020

• pretraining script

  ibeltagy

  committed
  95296ad

• wip

  ibeltagy

  committed
  325693e

Commits on Jul 17, 2020

• wip

  ibeltagy

  committed
  985acc9

• wip

  ibeltagy

  committed
  023dd78

• wip

  ibeltagy

  committed
  08230ac

• .

  ibeltagy

  committed
  fb65d57

• pad chunks or start next doc

  ibeltagy

  committed
  0e80cde

• todo

  ibeltagy

  committed
  6ca7d1b

• wip

  ibeltagy

  committed
  a2aa4f7

• wip

  ibeltagy

  committed
  62a69d5

Commits on Jul 18, 2020

• wip

  ibeltagy

  committed
  3e3a478

• wip

  ibeltagy

  committed
  3bc5354

• wip

  ibeltagy

  committed
  1a91024

• wip

  ibeltagy

  committed
  5fa21f2

• wip

  ibeltagy

  committed
  18eb003

• wip

  ibeltagy

  committed
  607e446

• wip

  ibeltagy

  committed
  d4659de

Commits on Jul 19, 2020

• wip

  ibeltagy

  committed
  c7c53cb

Commits on Jul 22, 2020

• wip

  ibeltagy

  committed
  0a07daf

• wip

  ibeltagy

  committed
  827576c

• wip

  ibeltagy

  committed
  5d0c8a2

• wip

  ibeltagy

  committed
  413258a

• Merge branch 'encoderdecoder' of github.com:allenai/longformer into encoderdecoder

  ibeltagy

  committed
  9b8a7d6

• tpu

  ibeltagy

  committed
  1a6498c

• wip

  ibeltagy

  committed
  3e82548

Commits on Jul 23, 2020

• wip

  ibeltagy

  committed
  adadd42

• pretraining script

  ibeltagy

  committed
  9e191a0

• wip

  ibeltagy

  committed
  9d18808

• wip

  ibeltagy

  committed
  6e24cee

• wip

  ibeltagy

  committed
  a2ab9b3

• wip

  ibeltagy

  committed
  e3f4ba9

• .

  ibeltagy

  committed
  f9e654b

• pad chunks or start next doc

  ibeltagy

  committed
  9c2646d

• todo

  ibeltagy

  committed
  433a2e2

• wip

  ibeltagy

  committed
  ec47270

• wip

  ibeltagy

  committed
  77e105d

• wip

  ibeltagy

  committed
  af08b5a

• wip

  ibeltagy

  committed
  d105023

• wip

  ibeltagy

  committed
  1183999

• wip

  ibeltagy

  committed
  20e8208

• wip

  ibeltagy

  committed
  224824d

• wip

  ibeltagy

  committed
  4a12730

• wip

  ibeltagy

  committed
  c936d24

• wip

  ibeltagy

  committed
  510801b

• wip

  ibeltagy

  committed
  9184b71

• wip

  ibeltagy

  committed
  4ae991a

• tpu

  ibeltagy

  committed
  aea2a98

• wip

  ibeltagy

  committed
  69b717a

• wip

  ibeltagy

  committed
  5f641c0

• wip

  ibeltagy

  committed
  e3ddeca

• Merge branch 'trainer' of github.com:allenai/longformer into trainer

  ibeltagy

  committed
  21c9e57

• wip

  ibeltagy

  committed
  00ce1e9

• wip

  ibeltagy

  committed
  56b9c6a

Commits on Jul 25, 2020

• wip

  ibeltagy

  committed
  8fca187

• wip

  ibeltagy

  committed
  9dd76b7

• wip

  ibeltagy

  committed
  d40983a

• wip

  ibeltagy

  committed
  f0f6a30

• Merge branch 'trainer' of github.com:allenai/longformer into trainer

  ibeltagy

  committed
  a6e37df

• wip

  ibeltagy

  committed
  9eb6fdf

• wip

  ibeltagy

  committed
  14b6074

• wip

  ibeltagy

  committed
  5b97bd6

• wip

  ibeltagy

  committed
  71d7a9d

• wip

  ibeltagy

  committed
  97a126d

• wip

  ibeltagy

  committed
  c873da2

Commits on Jul 28, 2020

• faster gradnorm

  ibeltagy

  committed
  d602869

• allow changing seqlen at runtime

  ibeltagy

  committed
  ffd06dd

Commits on Jul 30, 2020

• log and resume data preprocessing

  ibeltagy

  committed
  129a3f9

• multiprocessed preprocessing

  ibeltagy

  committed
  1c42f96

Commits on Aug 3, 2020

• wip

  ibeltagy

  committed
  c20264e

• Save this directory as a dataset and use it directly on a plain base image.

  meslater1030

  committed
  ff96351

Commits on Aug 6, 2020

• bug fix

  ibeltagy

  committed
  0557e24

Commits on Aug 7, 2020

• fix a bug with the mapping from longformerselfattention to bartselfattention

  ibeltagy

  committed
  6ae5051

• mem_profiler

  ibeltagy

  committed
  a1de977

Commits on Aug 12, 2020

• extend encoder only

  ibeltagy

  committed
  1bf6c7c

• upgrade triviaqa script to PLv0.8.5

  ibeltagy

  committed
  5b31f5e

Commits on Aug 13, 2020

• add roberta baseline

  ibeltagy

  committed
  405739e

Commits on Aug 17, 2020

• Merge branch 'mes/longformer-on-beaker-copy-all' into encoderdecoder

  ibeltagy

  committed
  162c22f

• triviaqa seq2seq + fix bart-base bug

  ibeltagy

  committed
  c132d4e

• Merge branch 'encoderdecoder' of github.com:allenai/longformer into encoderdecoder

  ibeltagy

  committed
  671aa72

Commits on Aug 23, 2020

• beaker

  ibeltagy

  committed
  d1349e9

Commits on Aug 25, 2020

• sliding_chunks_no_overlap (#100)

  ibeltagy

  authored
  5a2b9da

Commits on Aug 26, 2020

• seq2seq

  ibeltagy

  committed
  b15607b

Commits on Aug 27, 2020

• wip

  ibeltagy

  committed
  82741c3

• wip

  ibeltagy

  committed
  5c3a22a

Commits on Aug 28, 2020

• wip

  ibeltagy

  committed
  75aeb47

Commits on Sep 1, 2020

• Update README.md

  ibeltagy

  authored
  391d8de

Commits on Sep 2, 2020

• Merge branch 'master' into encoderdecoder

  ibeltagy

  committed
  3aa2f67

• summarization

  ibeltagy

  committed
  bf9e58a

Commits on Sep 3, 2020

• fix loading data

  ibeltagy

  committed
  0158125

• wip

  ibeltagy

  committed
  9fdef52

• wip

  ibeltagy

  committed
  cbb407d

• grad_ckpt + reqs + long

  ibeltagy

  committed
  eb34cc0

Commits on Sep 4, 2020

• ignore empty answers

  ibeltagy

  committed
  42481fd

• Merge branch 'encoderdecoder' of github.com:allenai/longformer into encoderdecoder

  ibeltagy

  committed
  79b9b0d

• attention dropout

  ibeltagy

  committed
  c6f2335

• Merge branch 'encoderdecoder' of github.com:allenai/longformer into encoderdecoder

  ibeltagy

  committed
  f5b9498

• model.generate takes a lot of memory. Set requires_grad=False

  ibeltagy

  committed
  f5a798d

Commits on Sep 5, 2020

• wip

  ibeltagy

  committed
  274d017

• wip

  ibeltagy

  committed
  5f765b9

Commits on Sep 6, 2020

• attention_mode

  ibeltagy

  committed
  5784aee

Commits on Sep 7, 2020

• wip

  ibeltagy

  committed
  b78384a

• Merge branch 'encoderdecoder' of github.com:allenai/longformer into encoderdecoder

  ibeltagy

  committed
  48c0344

• pegasus bug

  ibeltagy

  committed
  327b729

• run on cpu

  ibeltagy

  committed
  4944fb8

• readme

  ibeltagy

  committed
  36252c0

• readme

  ibeltagy

  committed
  281999f

Commits on Sep 9, 2020

• adafactor and label smoothing

  ibeltagy

  committed
  dee3daf

Commits on Sep 13, 2020

• add rougeLsum

  ibeltagy

  committed
  498ca04

• Merge branch 'encoderdecoder' of github.com:allenai/longformer into encoderdecoder

  ibeltagy

  committed
  d928dfb

Commits on Nov 12, 2020

• wip code for LongT5

  AkshitaB

  committed
  0f3875f

Commits on Nov 23, 2020

• naive code for smaller score matrix

  AkshitaB

  committed
  adc92ca

• commenting attn_mask

  AkshitaB

  committed
  02f9c3f

Commits on Nov 24, 2020

• fixing order and issue with inf

  AkshitaB

  committed
  de147de

Commits on Dec 8, 2020

• fixing compute_bias

  AkshitaB

  committed
  1ba5286