Code

🤝 Mixture of judges#2159

Code

Merged

qgallouedec merged 46 commits intohuggingface:mainhuggingface/trl:mainfrom gaetanlop:cgpo_mixture_of_judgesgaetanlop/trl:cgpo_mixture_of_judgesCopy head branch name to clipboardNov 18, 2024

+149-7

Commits

Commits on Oct 3, 2024

• base judge

  gaetanlop

  committed
  013aae4

  Copy full SHA for 013aae4

• adding mixture of judges

  gaetanlop

  committed
  0ea5a48

  Copy full SHA for 0ea5a48

• update doc

  gaetanlop

  committed
  517cfb0

  Copy full SHA for 517cfb0

• update doc

  gaetanlop

  committed
  9e5ed12

  Copy full SHA for 9e5ed12

• formatting

  gaetanlop

  committed
  3406e53

  Copy full SHA for 3406e53

• fix small typo in doc

  gaetanlop

  committed
  568d2b9

  Copy full SHA for 568d2b9

• fix randomcontraintjudge

  gaetanlop

  committed
  466292e

  Copy full SHA for 466292e

Commits on Oct 4, 2024

• Merge branch 'main' into cgpo_mixture_of_judges

  qgallouedec

  authored
  a3d90df

  Copy full SHA for a3d90df

• replace arxiv by hf papers

  Show description for 3f0b8b0
  

  gaetanlop

  and

  qgallouedec

  authored
  3f0b8b0

  Copy full SHA for 3f0b8b0

• formatting

  Show description for 8995ab4
  

  gaetanlop

  and

  qgallouedec

  authored
  8995ab4

  Copy full SHA for 8995ab4

• Merge branch 'main' into cgpo_mixture_of_judges

  gaetanlop

  authored
  896259e

  Copy full SHA for 896259e

• fix naming in __init__

  gaetanlop

  committed
  ef1feb0

  Copy full SHA for ef1feb0

• run precommi

  gaetanlop

  committed
  3da4a06

  Copy full SHA for 3da4a06

Commits on Oct 7, 2024

• adding gold answers to judges

  gaetanlop

  committed
  765768b

  Copy full SHA for 765768b

• cgpo llm judges

  gaetanlop

  committed
  8aaaaa1

  Copy full SHA for 8aaaaa1

• fix init

  gaetanlop

  committed
  cfc84ed

  Copy full SHA for cfc84ed

• Merge branch 'main' into cgpo_mixture_of_judges

  gaetanlop

  authored
  a1e8eeb

  Copy full SHA for a1e8eeb

• output type

  gaetanlop

  committed
  6898285

  Copy full SHA for 6898285

• adjust booleans in test

  gaetanlop

  committed
  f5639a1

  Copy full SHA for f5639a1

• adapt moj doc

  gaetanlop

  committed
  289b855

  Copy full SHA for 289b855

Commits on Oct 9, 2024

• Merge branch 'main' into cgpo_mixture_of_judges

  gaetanlop

  authored
  308e743

  Copy full SHA for 308e743

Commits on Oct 11, 2024

• Merge branch 'main' into cgpo_mixture_of_judges

  gaetanlop

  authored
  2c6de87

  Copy full SHA for 2c6de87

• renaming and removing factuality and safety judges

  gaetanlop

  committed
  dedc859

  Copy full SHA for dedc859

• fix typo in import

  gaetanlop

  committed
  ba0fffb

  Copy full SHA for ba0fffb

• fix small typo in naming

  gaetanlop

  committed
  226de82

  Copy full SHA for 226de82

Commits on Oct 14, 2024

• Merge branch 'main' into cgpo_mixture_of_judges

  gaetanlop

  authored
  5626cd4

  Copy full SHA for 5626cd4

• formatting

  gaetanlop

  committed
  567b798

  Copy full SHA for 567b798

Commits on Oct 21, 2024

• Merge branch 'main' into cgpo_mixture_of_judges

  gaetanlop

  authored
  1c33494

  Copy full SHA for 1c33494

Commits on Oct 24, 2024

• Merge branch 'main' into cgpo_mixture_of_judges

  gaetanlop

  authored
  64c9de8

  Copy full SHA for 64c9de8

• Update trl/trainer/judges.py

  Show description for 559cd1b
  

  gaetanlop

  and

  qgallouedec

  authored
  559cd1b

  Copy full SHA for 559cd1b

Commits on Oct 25, 2024

• update parameter name

  gaetanlop

  committed
  2c29ef5

  Copy full SHA for 2c29ef5

• update tests

  gaetanlop

  committed
  bd1bed8

  Copy full SHA for bd1bed8

• update doc

  gaetanlop

  committed
  21e3ccd

  Copy full SHA for 21e3ccd

Commits on Oct 29, 2024

• Merge branch 'main' into cgpo_mixture_of_judges

  gaetanlop

  authored
  43d6cca

  Copy full SHA for 43d6cca

• Update trl/trainer/judges.py

  Show description for 9eca0f8
  

  gaetanlop

  and

  qgallouedec

  authored
  9eca0f8

  Copy full SHA for 9eca0f8

• Update doc

  Show description for d5b32f0
  

  gaetanlop

  and

  qgallouedec

  authored
  d5b32f0

  Copy full SHA for d5b32f0

• fix alltruejudge type

  gaetanlop

  committed
  ac88c63

  Copy full SHA for ac88c63

Commits on Nov 4, 2024

• Merge branch 'main' into cgpo_mixture_of_judges

  gaetanlop

  authored
  999154b

  Copy full SHA for 999154b

Commits on Nov 15, 2024

• Merge branch 'main' into cgpo_mixture_of_judges

  qgallouedec

  authored
  6d7c8bc

  Copy full SHA for 6d7c8bc

• Refactor judge variable names and update test names

  qgallouedec

  committed
  8b36bc7

  Copy full SHA for 8b36bc7

• Clarify judgment logic

  qgallouedec

  committed
  30f3ca8

  Copy full SHA for 30f3ca8

• Merge branch 'cgpo_mixture_of_judges' of https://github.com/gaetanlop/trl into cgpo_mixture_of_judges

  qgallouedec

  committed
  60af3c4

  Copy full SHA for 60af3c4

• Fix invalid binary judgment check in AllTrueJudge class

  qgallouedec

  committed
  fcfcd14

  Copy full SHA for fcfcd14

• Fix invalid binary judgment check in AllTrueJudge class

  qgallouedec

  committed
  7e7f9e7

  Copy full SHA for 7e7f9e7

Commits on Nov 18, 2024

• Merge branch 'main' into cgpo_mixture_of_judges

  qgallouedec

  authored
  294588f

  Copy full SHA for 294588f

• Merge branch 'main' into cgpo_mixture_of_judges

  qgallouedec

  authored
  f049351

  Copy full SHA for f049351