add max_seq_len checks

fix config

test emb

Upload model.safetensors with huggingface_hub

adjust tokenizer config

init