diff --git a/examples/lightrag_lmdeploy_demo.py b/examples/lightrag_lmdeploy_demo.py index ea7ace0e..aeb96f71 100644 --- a/examples/lightrag_lmdeploy_demo.py +++ b/examples/lightrag_lmdeploy_demo.py @@ -10,10 +10,11 @@ WORKING_DIR = "./dickens" if not os.path.exists(WORKING_DIR): os.mkdir(WORKING_DIR) + async def lmdeploy_model_complete( prompt=None, system_prompt=None, history_messages=[], **kwargs ) -> str: - model_name = kwargs["hashing_kv"].global_config["llm_model_name"] + model_name = kwargs["hashing_kv"].global_config["llm_model_name"] return await lmdeploy_model_if_cache( model_name, prompt, @@ -23,7 +24,7 @@ async def lmdeploy_model_complete( ## or model_name is a pytorch model on huggingface.co, ## you can refer to https://github.com/InternLM/lmdeploy/blob/main/lmdeploy/model.py ## for a list of chat_template available in lmdeploy. - chat_template = "llama3", + chat_template="llama3", # model_format ='awq', # if you are using awq quantization model. # quant_policy=8, # if you want to use online kv cache, 4=kv int4, 8=kv int8. **kwargs, @@ -33,7 +34,7 @@ async def lmdeploy_model_complete( rag = LightRAG( working_dir=WORKING_DIR, llm_model_func=lmdeploy_model_complete, - llm_model_name="meta-llama/Llama-3.1-8B-Instruct", # please use definite path for local model + llm_model_name="meta-llama/Llama-3.1-8B-Instruct", # please use definite path for local model embedding_func=EmbeddingFunc( embedding_dim=384, max_token_size=5000,