Adding `safetensors` variant of this model

Added training logs

Saving weights and logs of step 801

Saving tokenizer

initial commit