Add custom JeevesTokenizer (wraps SentencePiece directly), fix config

Remove broken tokenizer.json (wrong token IDs)

Fix: tie_word_embeddings=false

Fix tokenizer: update special_tokens_map.json for transformers 5.x compat

Fix tokenizer: update tokenizer_config.json for transformers 5.x compat

Upload Jeeves-Small-100M

Fix: set tie_word_embeddings=false (weights stored separately in safetensors)

Fix tokenizer: update special_tokens_map.json for transformers 5.x compat

Fix tokenizer: update tokenizer_config.json for transformers 5.x compat

Fix tokenizer: update tokenizer.json for transformers 5.x compat

Update README.md

Upload Jeeves-Small-100M

initial commit