* ChatGLM3-6B LoRA Fine-tuning Demo * refine * refine * add 2-card deepspeed * refine format * add mpi4py and deepspeed install