Commit History

Autor SHA1 Mensaxe Data
  кѳѳsнī 55b664f46f Enabling model parallelism (training 30b on 2x 3090s and beyond) (#131) %!s(int64=3) %!d(string=hai) anos
  Eric Wang 3b79ea4029 256 -> 512 -> 256 %!s(int64=3) %!d(string=hai) anos
  Eric Wang 804d22ad43 remove asserts %!s(int64=3) %!d(string=hai) anos
  Angainor Development 69b9d9ea8b Fix a warning (#186) %!s(int64=3) %!d(string=hai) anos
  Eric J. Wang dbd04f3560 Fix linters (#185) %!s(int64=3) %!d(string=hai) anos
  NanoCode012 69b31e0fed Feat: Add wandb (#168) %!s(int64=3) %!d(string=hai) anos
  claysauruswrecks 1310547f9f Add HF dataset loading, add linters, pyproject.toml (#175) %!s(int64=3) %!d(string=hai) anos
  Angainor Development 9d6b822019 Avoid a deprecation warning (#181) %!s(int64=3) %!d(string=hai) anos
  Eric Wang 683810b4a1 Print warning on checkpoint not found %!s(int64=3) %!d(string=hai) anos
  Eric Wang da6b427a08 resume_from_checkpoint %!s(int64=3) %!d(string=hai) anos
  Eric Wang b948f892ba restore default settings %!s(int64=3) %!d(string=hai) anos
  Eric J. Wang 5fa807d106 Use CLI arguments (#159) %!s(int64=3) %!d(string=hai) anos
  Eric J. Wang af30df1999 Unified tokenizer update PR (#146) %!s(int64=3) %!d(string=hai) anos
  Eric Wang 72aabcb5a4 Remove LLaMA download code, as a precaution %!s(int64=3) %!d(string=hai) anos
  Eric Wang b12c3b90f8 Unwind input masking to avoid confusion %!s(int64=3) %!d(string=hai) anos
  Eric Wang 7fb06c6c22 Revert "Mask out prompt tokens for real" %!s(int64=3) %!d(string=hai) anos
  Eric Wang 2204a71505 set EPOCHS back to 3 %!s(int64=3) %!d(string=hai) anos
  Eric Wang 4a712d4d8e Mask out prompt tokens for real %!s(int64=3) %!d(string=hai) anos
  Eric Wang fac53721a2 masking bugfix %!s(int64=3) %!d(string=hai) anos
  Kohaku-Blueleaf b5a1a0bca7 Add support for valid set size 0 (#83) %!s(int64=3) %!d(string=hai) anos
  Kohaku-Blueleaf 0af44f0262 Add option for output dir (#84) %!s(int64=3) %!d(string=hai) anos
  Kohaku-Blueleaf 450206caaf Fix torch.compile call on windows (#81) %!s(int64=3) %!d(string=hai) anos
  Eric Wang cfad895aa1 mask prompt in loss %!s(int64=3) %!d(string=hai) anos
  Kakigōri Maker 9dab7ba438 add multi-gpu support (ddp) (#54) %!s(int64=3) %!d(string=hai) anos
  Eric Wang f7044049ab dataset cleaning, visualizations %!s(int64=3) %!d(string=hai) anos
  Eric Wang 35029da078 Validation set %!s(int64=3) %!d(string=hai) anos
  Eric Wang 5f6614e6fc Catch outdated installs %!s(int64=3) %!d(string=hai) anos
  andreas.echavez 1862976b33 Update alpaca-lora to use transformers main branch %!s(int64=3) %!d(string=hai) anos
  Eric Wang 2fa1c66388 repair tokenization logic, again %!s(int64=3) %!d(string=hai) anos
  Eric Wang 024dde7dab Revert "fix <eos> tokenization" %!s(int64=3) %!d(string=hai) anos