LoRE / scripts

Commit History

  1. batch 3

    85d731b
  2. batch size 2

    d55ddf7
  3. batch 4

    0d64997
  4. comma

    d2653ee
  5. 16batch

    7d3ca95
  6. train agaaa

    45d6e50
  7. train aga

    580eff8
  8. fix train

    1f3825f

  1. debugging missing grad

    325d2d0
  2. update training script

    f9596a0

  1. reorder small experts

    7050cb6
  2. unfreeze only gate and experts

    356573e
  3. 1 batch size

    6b0e19d

  1. batch size back to 2

    14b2125
  2. batch size 3 lol

    f078b84
  3. batch size to 4

    5f8bb1e
  4. batch size 8

    9fc70e4
  5. modify batch and fix tensor issue

    2a594f6
  6. tokenize fn

    5c05368
  7. fix

    52bdc02
  8. add

    d6ffab2
  9. tokenize function

    1e0b293
  10. debugs

    6d21fca
  11. key value

    7ab89f2
  12. restore

    7abbd62
  13. claude attempt 2 dataset

    3db4e2e
  14. cache diagnostics

    dd2e997
  15. sanity

    8e88ea1
  16. claudeattempt dataset

    5b01886
  17. alternative dataset load

    d7f70e5
  18. dataset keep in memory

    1182794
  19. ignore mismatches

    e039ec3
  20. fix import

    78b85e8
  21. overhaul

    c4785c5
  22. reset modeling file

    36acce3

  1. attempt fix and more prints

    a82f934
  2. init expanded model after config change

    438a56a

  1. config shenanigans

    72dfb47
  2. debug to train script

    e2fe765
  3. changes without config

    b41d150
  4. update training script

    582cd6b
  5. restore import name

    e3a54b7
  6. handle base and product architecture differences

    b63994d
  7. import path

    50a970c
  8. import file name

    34f196f
  9. initial stuff

    48f0e60

  1. init no routing

    9d66be3