llama.cpp

d0330fd7 - py : add capabiliy to convert from ggml back to torch or hf format for further consumption/training/finetuning (#403)