Adding `safetensors` variant of this model

upload flax model

allow flax

continued training

training with tokenizer on

edit tokenizer

initial commit

initial commit