Create LLM_low_bit_optimize.py

LLM_low_bit_optimize.py

@@ -0,0 +1,9 @@
+from ipex_llm.transformers import AutoModelForCausalLM
+from transformers import LlamaTokenizer
+
+
+llm = AutoModelForCausalLM.from_pretrained("checkpoints\\Llama-2-7b-chat-hf",load_in_low_bit="sym_int4")
+llm.save_low_bit("checkpoints\\Llama-2-7b-chat-hf-INT4")
+
+tokenizer = LlamaTokenizer.from_pretrained("checkpoints\\Llama-2-7b-chat-hf\\")
+tokenizer.save_pretrained("checkpoints\\Llama-2-7b-chat-hf-INT4")