add fine-tuned model trained on multilingual conversations based on decapoda-research/llama-13b-hf model

add tokenizer_config.json file from original model