-
Add BPE tokenizer (needed for Colab training notebooks)
-
Add symbio_model.py for Colab notebook imports
-
Add 2MB val text sample for Gemma/HF tokenizer notebooks
-
Add 20MB raw text sample for Gemma/HF tokenizer notebooks
-
Upload README.md with huggingface_hub
-
Distillation test winner (scratch, PPL=43.9, 5M params)
-
Upload merges.txt with huggingface_hub
-
Upload vocab.json with huggingface_hub
-
Upload best checkpoint (val_loss=3.5599)
-
Upload SymbioGPT-10M teacher (val_ppl=35.3, 13400 steps, A100)
-
Add curated training tokens (266M tokens, Chinchilla-optimal)
-
Add validation tokens (72M tokens)
-
initial commit