Datasheet: add effective lexicon-usage accounting (corpus-intersecting slice; 9.90M < 10M under strict counting)

Datasheet correction: MorphyNet used untruncated in the submission pipeline (160k cutoff is analysis/Tier-2 only)

Add data statement (datasheet): cleaned strict-small corpus, budget accounting, tokenizer resources, checkpoint index

Fix train_meta.json: correct to m20-ms (was mislabeled v5pm-ms/mlm-0.5)

Correct model card for v5: name, params (135M), recipe (Muon+MNTP@0.20), BLiMP 72.52

Upload model.safetensors with huggingface_hub

Upload tokenization_ourmorph.py with huggingface_hub

Upload our_tokenizer.py with huggingface_hub

Upload folder using huggingface_hub

initial commit