Code

fix pretraining_ on odd datasets#1463

Code

Merged

winglian merged 9 commits intoaxolotl-ai-cloud:mainaxolotl-ai-cloud/axolotl:mainfrom mapmeld:pretrain_cfgCopy head branch name to clipboardApr 2, 2024

+25-7

Commits

Commits on Mar 30, 2024

• can configure name of split of pretraining dataset

  mapmeld

  committed
  27dc6e2

• streaming data and dataset map

  mapmeld

  committed
  d446ee1

• text column customized

  mapmeld

  committed
  d3ec1b9

• allow text_column to be set in pretrain

  mapmeld

  committed
  86e3a2b

• pretrain type

  mapmeld

  committed
  4d53947

• load a bit of the dataset

  mapmeld

  committed
  577edcb

• fix dataset where splits have separate configs

  mapmeld

  committed
  345cbc8

• ok name param here is the config

  mapmeld

  committed
  cf1843b

• whitespace

  mapmeld

  committed
  f95b7e0