Code

Llama/GPTNeoX: add RoPE scaling #24653

Code

Merged

gante merged 9 commits intohuggingface:mainhuggingface/transformers:mainfrom gante:pos_emb_scalinggante/transformers:pos_emb_scalingCopy head branch name to clipboardJul 13, 2023

+484-67

Commits

Commits on Jul 4, 2023

• add rope_scaling

  gante

  committed
  e7d743f

  Copy full SHA for e7d743f

Commits on Jul 11, 2023

• tmp commit

  gante

  committed
  11a25da

  Copy full SHA for 11a25da

Commits on Jul 12, 2023

• add gptneox

  gante

  committed
  022c977

  Copy full SHA for 022c977

• add tests

  gante

  committed
  c12aaf1

  Copy full SHA for c12aaf1

• Merge branch 'main' into pos_emb_scaling

  gante

  authored
  d7e7636

  Copy full SHA for d7e7636

• GPTNeoX can now handle long inputs, so the pipeline test was wrong

  gante

  committed
  8c52524

  Copy full SHA for 8c52524

Commits on Jul 13, 2023

• Update src/transformers/models/open_llama/configuration_open_llama.py

  Show description for c7161b4
  

  gante

  and

  amyeroberts

  authored
  c7161b4

  Copy full SHA for c7161b4

• remove ntk

  gante

  committed
  7c02d92

  Copy full SHA for 7c02d92

• remove redundant validation

  gante

  committed
  f385927

  Copy full SHA for f385927