Code

Make the TF dummies even smaller#24071

Code

Merged

Rocketknight1 merged 6 commits intomainhuggingface/transformers:mainfrom even_lower_tf_dummy_memoryhuggingface/transformers:even_lower_tf_dummy_memoryCopy head branch name to clipboardJun 7, 2023

+12-23

Commits

Commits on Jun 7, 2023

• Let's see if we can use the smallest possible dummies

  Rocketknight1

  committed
  b0b4d1b

  Copy full SHA for b0b4d1b

• Make GPT-2's dummies a little longer

  Rocketknight1

  committed
  8d6840f

  Copy full SHA for 8d6840f

• Just use (1,2) as the default shape

  Rocketknight1

  committed
  47e3224

  Copy full SHA for 47e3224

• Update other dummies in sync

  Rocketknight1

  committed
  a8d2f4c

  Copy full SHA for a8d2f4c

• Correct imports for Keras 2.13

  Rocketknight1

  committed
  e593919

  Copy full SHA for e593919

• Shrink the Wav2Vec2 dummies

  Rocketknight1

  committed
  c7d9962

  Copy full SHA for c7d9962