Commit History

  1. fix embedding & logits order, add tokenizer

    819c65a

  1. Step 10205, loss 2.184

    44f1603

  1. Step 10200, loss 2.158

    96d524b
  2. Step 10195, loss 2.202

    197c1bc
  3. Step 10190, loss 2.132

    85ed573
  4. Step 10185, loss 2.190

    e4b5632
  5. Step 10180, loss 2.101

    e92bc71
  6. Step 10175, loss 2.105

    00a46e8
  7. Step 10170, loss 2.160

    2d991d1
  8. Step 10165, loss 2.196

    29f40fd
  9. Step 10160, loss 2.178

    b1a7027
  10. Step 10155, loss 2.182

    431f73a
  11. Step 10145, loss 2.166

    f10997c
  12. Step 10130, loss 2.190

    c37b29b
  13. Step 10115, loss 2.223

    0b88971
  14. Step 10100, loss 2.215

    3024318
  15. Step 10085, loss 2.156

    bef9ef7
  16. Step 10070, loss 2.245

    89a142d
  17. Step 10055, loss 2.206

    309ba5c
  18. Step 10040, loss 2.170

    02aed64
  19. Step 10025, loss 2.211

    a7ad92b
  20. Step 10010, loss 2.222

    fa2694b
  21. Step 9995, loss 2.152

    06f8a62

  1. Step 9980, loss 2.174

    a25afe5
  2. Step 9965, loss 2.115

    facaa25
  3. Step 9950, loss 2.206

    0df23d6
  4. Step 9935, loss 2.123

    d25f9be
  5. Step 9920, loss 2.012

    9deb687
  6. Step 9905, loss 2.162

    aa0649a
  7. Step 9890, loss 2.205

    5aa80e9
  8. Step 9875, loss 2.245

    fae1123
  9. Step 9860, loss 2.139

    97e7344
  10. Step 9845, loss 2.178

    87fcaf4
  11. Step 9830, loss 2.196

    ff02282
  12. Step 9815, loss 2.203

    f81bde5
  13. Step 9800, loss 2.189

    524cfda
  14. Step 9785, loss 2.148

    9a45b7d
  15. Step 9770, loss 2.318

    fda23c6
  16. Step 9755, loss 2.174

    8ac2f43
  17. Step 9740, loss 2.286

    83dc916
  18. Step 9725, loss 2.244

    8f8fb9a
  19. Step 9710, loss 2.228

    30a8639
  20. Step 9695, loss 2.210

    0558311
  21. Step 9680, loss 2.221

    e2cad3f

  1. Step 9665, loss 2.203

    552f3c3
  2. Step 9650, loss 2.212

    d4f8b90
  3. Step 9635, loss 2.263

    46c8a86
  4. Step 9620, loss 2.290

    0a7e32c
  5. Step 9605, loss 2.280

    ec25517
  6. Step 9590, loss 2.226

    11f9020