Commit History

  1. Add BPE tokenizer (needed for Colab training notebooks)

    Verified signature
    215b74b

  1. Add symbio_model.py for Colab notebook imports

    Verified signature
    d968181
  2. Add 2MB val text sample for Gemma/HF tokenizer notebooks

    Verified signature
    588ff7f
  3. Add 20MB raw text sample for Gemma/HF tokenizer notebooks

    Verified signature
    d9422a5

  1. Upload README.md with huggingface_hub

    Verified signature
    4bf22dd
  2. Distillation test winner (scratch, PPL=43.9, 5M params)

    Verified signature
    49d0c4c
  3. Upload merges.txt with huggingface_hub

    Verified signature
    b7fe905
  4. Upload vocab.json with huggingface_hub

    Verified signature
    912c171
  5. Upload best checkpoint (val_loss=3.5599)

    Verified signature
    69e2d95
  6. Upload SymbioGPT-10M teacher (val_ppl=35.3, 13400 steps, A100)

    Verified signature
    06b4943
  7. Add curated training tokens (266M tokens, Chinchilla-optimal)

    Verified signature
    ab3f5b8
  8. Add validation tokens (72M tokens)

    Verified signature
    ff30a4a
  9. initial commit

    Verified signature
    4c8061f