Finetuning Llama models using converted weights.

[AravAct]

Thanks for this amazing repo. This is the first repo using which I am successfully able to run 13B model (using 2 V100 GPU).
Are there any plans to add finetune capability and also add lower quantization (int8, int4).