JayLZhou
Query bugs
Hi, I'm sorry to bother you.
While testing query results using main.py, I encountered an issue. Specifically, when the number of questions exceeds two, I receive an error message stating, "Event loop is closed."
RuntimeError: Event loop is closed
Traceback (most recent call last):
File "/home/lcy/GraphRAG-exp/main.py", line 81, in conda run python /home/lcy/GraphRAG-exp/main.py -opt Option/Method/LightRAG.yaml -dataset_name datasets/test failed. (See above for error)
I am using the LightRAG.yaml file for the configuration settings. Could you help me?
This is because, your seed is none. That is, no relationships are retrieved. Please show me your question maybe
Thanks for your reply! I met this issue when the program process the second question in multihop-rag dataset, i.e., {"question":"Which individual is implicated in both inflating the value of a Manhattan apartment to a figure not yet achieved in New York City's real estate history, according to 'Fortune', and is also accused of adjusting this apartment's valuation to compensate for a loss in another asset's worth, as reported by 'The Age'?","answer":"Donald Trump","label":"Donald Trump"}
I think I have build the graph successfully. The output is as follows. #################################################################################################### 2025-03-30 06:23:40.535 | INFO | Core.Chunk.DocChunk:build_chunks:27 - Starting chunk the given documents 2025-03-30 06:23:40.813 | INFO | Core.Storage.ChunkKVStorage:_persist:105 - Writing data into ./datasets/multihop-rag/rkg_graph/chunk_storage_chunk_data_idx.pkl and ./datasets/multihop-rag/rkg_graph/chunk_storage_chunk_data_key.pkl 2025-03-30 06:23:40.836 | INFO | Core.Chunk.DocChunk:build_chunks:74 - ✅ Finished the chunking stage 2025-03-30 06:23:40.845 | INFO | Core.GraphRAG:_update_costs_info:205 - Chunking stage cost: Total prompt token: 0, Total completeion token: 0, Total cost: 0 2025-03-30 06:23:40.846 | INFO | Core.GraphRAG:_update_costs_info:207 - Chunking time(s): 0.31 2025-03-30 06:23:40.846 | INFO | Core.Graph.BaseGraph:build_graph:41 - Starting build graph for the given documents 2025-03-30 06:23:40.846 | INFO | Core.Storage.NetworkXStorage:load_nx_graph:27 - Attempting to load the graph from: ./datasets/multihop-rag/rkg_graph/graph_storage_nx_data.graphml 2025-03-30 06:23:41.341 | INFO | Core.Storage.NetworkXStorage:load_nx_graph:31 - Successfully loaded graph from: ./datasets/multihop-rag/rkg_graph/graph_storage_nx_data.graphml with 14551 nodes and 7442 edges 2025-03-30 06:23:41.341 | INFO | Core.Graph.BaseGraph:build_graph:50 - ✅ Finished the graph building stage 2025-03-30 06:23:41.341 | INFO | Core.GraphRAG:_update_costs_info:205 - Build Graph stage cost: Total prompt token: 0, Total completeion token: 0, Total cost: 0 2025-03-30 06:23:41.341 | INFO | Core.GraphRAG:_update_costs_info:207 - Build Graph time(s): 0.50 2025-03-30 06:23:41.622 | INFO | Core.Index.BaseIndex:build_index:13 - Starting insert elements of the given graph into vector database 2025-03-30 06:23:41.622 | INFO | Core.Index.BaseIndex:build_index:17 - Loading index from the file ./datasets/multihop-rag/rkg_graph/entities_vdb 2025-03-30 06:24:34.659 | INFO | Core.Index.BaseIndex:build_index:29 - ✅ Finished starting insert entities of the given graph into vector database 2025-03-30 06:24:34.717 | INFO | Core.Index.BaseIndex:build_index:13 - Starting insert elements of the given graph into vector database 2025-03-30 06:24:34.717 | INFO | Core.Index.BaseIndex:build_index:17 - Loading index from the file ./datasets/multihop-rag/rkg_graph/relations_vdb 2025-03-30 06:25:02.264 | INFO | Core.Index.BaseIndex:build_index:29 - ✅ Finished starting insert entities of the given graph into vector database 2025-03-30 06:25:02.265 | INFO | Core.GraphRAG:_update_costs_info:205 - Index Building stage cost: Total prompt token: 0, Total completeion token: 0, Total cost: 0 2025-03-30 06:25:02.265 | INFO | Core.GraphRAG:_update_costs_info:207 - Index Building time(s): 80.92 2025-03-30 06:25:02.265 | INFO | Core.GraphRAG:_build_retriever_context:176 - Building retriever context for the current execution 0 2025-03-30 06:25:06.414 | INFO | Core.Common.CostManager:update_cost:136 - prompt_tokens: 411, completion_tokens: 55 2025-03-30 06:25:07.768 | INFO | Core.Query.BasicQuery:_retrieve_relevant_contexts_global_keywords:172 - Global query uses 1 entities, 5 relations, 3 text units 2025-03-30 06:25:10.599 | INFO | Core.Common.CostManager:update_cost:136 - prompt_tokens: 411, completion_tokens: 55 2025-03-30 06:25:10.976 | INFO | Core.Query.BasicQuery:_retrieve_relevant_contexts_local:75 - Using 5 entities, 42 relations, 4 text units 2025-03-30 06:25:13.807 | INFO | Core.Common.CostManager:update_cost:136 - prompt_tokens: 411, completion_tokens: 55 2025-03-30 06:25:14.017 | INFO | Core.Query.BasicQuery:_retrieve_relevant_contexts_global_keywords:172 - Global query uses 1 entities, 5 relations, 3 text units 2025-03-30 06:25:48.506 | INFO | Core.Common.CostManager:update_cost:136 - prompt_tokens: 8892, completion_tokens: 83 1 2025-03-30 06:25:52.894 | INFO | Core.Common.CostManager:update_cost:136 - prompt_tokens: 432, completion_tokens: 50 2025-03-30 06:25:52.895 | ERROR | Core.Retriever.RelationshipRetriever:_find_relevant_relations_vdb:48 - Failed to find relevant relationships: Event loop is closed Traceback (most recent call last):
########################
My setting file for LightRAG is as follow.
################################# Working settings #################################
# Basic Config
use_entities_vdb: True
use_relations_vdb: True # Only set True for LightRAG
use_subgraphs_vdb: False
llm_model_max_token_size: 32768
use_entity_link_chunk: False # Only set True for HippoRAG and FastGraphRAG
enable_graph_augmentation: False
# Data
index_name: rkg_graph
vdb_type: vector # vector/colbert
# Chunk Config
chunk:
chunk_token_size: 1200
chunk_overlap_token_size: 100
token_model: gpt-3.5-turbo
chunk_method: chunking_by_token_size
# Graph Config
graph:
# enable LightRAG
enable_edge_keywords: True
graph_type: rkg_graph # rkg_graph/er_graph/tree_graph/passage_graph
force: False
# Building graph
extract_two_step: True
max_gleaning: 1
enable_entity_description: True
enable_entity_type: False
enable_edge_description: True
enable_edge_name: True
# Retrieval Config
retriever:
query_type: basic
enable_local: False
use_entity_similarity_for_ppr: True
top_k_entity_for_ppr: 8
node_specificity: True
damping: 0.1
top_k: 5
query:
query_type: qa
only_need_context: False
enable_hybrid_query: True
augmentation_ppr: True
response_type: Multiple Paragraphs
level: 2
community_information: True
retrieve_top_k: 20
# naive search
naive_max_token_for_text_unit: 12000
# local search
local_max_token_for_text_unit: 4000 # 12000 * 0.33
max_token_for_text_unit: 4000
use_keywords: True
use_community: False
entities_max_tokens: 2000
relationships_max_tokens: 2000
max_token_for_local_context: 4800 # 12000 * 0.4
local_max_token_for_community_report: 3200 # 12000 * 0.27
local_community_single_one: False
# global search
use_global_query: True
global_min_community_rating: 0
global_max_consider_community: 512
global_max_token_for_community_report: 16384
max_token_for_global_context: 4000
global_special_community_map_llm_kwargs: {"response_format": {"type": "json_object"}}
# For IR-COT
max_ir_steps: 2
Here is the Config2.yaml
################################# Config2 settings #################################
# llm setting for the ollama
llm:
api_type: "open_llm" # open_llm or openai
base_url: 'http://localhost:11434/v1'
# model: "qwen2.5:14b"
model: "llama3.1:latest"
api_key: "sk-XXXXXXXXXXXXXXXX"
embedding:
api_type: "ollama" # hf/ollama/openai.
base_url: 'http://localhost:11434'
api_key: "YOUR_API_KEY"
model: "nomic-embed-text:latest"
cache_dir: "/embedding_cache/" # Cache directory for embedding models
dimensions: 1024
max_token_size: 8102
embed_batch_size: 128
embedding_func_max_async: 16
data_root: "./" # Root directory for data
working_dir: ./ # Result directory for the experiment
#exp_name: "light_rag_exp" # Experiment name
exp_name: "LGraphRAG_exp"
#
Hi. i want to know your node_datas is null? Sorry for the latter response.
Thank you for your reply. Based on my tests, I can confirm that node_datas is not null when I met the problem. The edge_datas in function _retrieve_relevant_contexts_global_keywords is none.
Additionally, I’ve observed that when I run the questions one by one in debug mode, it is ok. This leads me to suspect that the problem might be related to asynchronous programming. Since I’m not familiar with asynchronous programming, so I’m unable to solve this issue.
Syr for the latter response, we will check how to support the batch updating for the question evualting. A.S.A.P
The issue isn't fully resolved yet. I attempted to modify the def wrapper_query(query_dataset, digimon, result_dir) function in main.py to be asynchronous, replacing res = asyncio.run(digimon.query(query["question"])) with res = await digimon.query(query["question"]). This allows the program to process more than two questions, but I still encounter a bug when processing the 104th question in the multihop-rag dataset.
Thank you！It is helpful for me！ Do you think it's possible that there's an issue with the environment setup?A version conflict occurred when I was configuring the requirements.txt.
Hi, we have updated the new requirement. Please see: https://github.com/JayLZhou/GraphRAG/blob/master/experiment.yml