[Bug]: REST V1 and V2: empty response when one of the candidate has nan distance

[dumbPy]

Is there an existing issue for this?

• [X] I have searched the existing issues

Environment

- Milvus version: v2.4.1
- Deployment mode(standalone or cluster): standalone
- SDK version(e.g. pymilvus v2.0.0rc2): 2.4.1
- OS(Ubuntu or CentOS): macOS

Current Behavior

When one of the candidate has nan distance, the rest api returns empty string as response.
Issue exists in both v1 and v2 rest apis.
This doesn't seem to be the issue with pymilvus

In below image, the pymilvus client returns valid objects while the rest api works fine till "limit":4 and returns empty string for any further limit as the distance reduces to nan.

Not sure how to reproduce this since the embeddings I am using here are not something I can share. I am using InnerProduct metric if that matters, and the vectors are of 300 dimension.

Expected Behavior

No response

Steps To Reproduce

No response

Milvus Log

No response

Anything else?

No response

[xiaofan-luan]

/assign @PowderLi please help on that

[yanliang567]

/assign @PowderLi /unassign

[yanliang567]

/assign @zhuwenxing please keep an eye on this, thx

[zhuwenxing]

The weird thing about this issue is why the distance appears as "nan".

[dumbPy]

The entities in question have zero vectors and that seem to trigger the nan. It should be 0 instead of nan.

[zhuwenxing]

I tried to insert many all zero vectors and search on them, the distance is displayed well as 0, and the response also was not empty.

[2024-05-11 11:12:51 - DEBUG - urllib3.connectionpool]: http://10.104.1.35:19530 "POST /v2/vectordb/collections/create HTTP/1.1" 200 22 (connectionpool.py:473)
[2024-05-11 11:12:51 - DEBUG - ci_test]: 
method: post, 
url: http://10.104.1.35:19530/v2/vectordb/collections/create, 
cost time: 0.14714789390563965, 
header: {'Content-Type': 'application/json', 'Authorization': 'Bearer root:Milvus', 'RequestId': '5a07f456-0f44-11ef-9c92-acde48001122'}, 
payload: {"collectionName": "test_collection_2024_05_11_11_12_50_440059vwiViLkk", "schema": {"autoId": true, "enableDynamicField": true, "fields": [{"fieldName": "book_id", "dataType": "Int64", "isPrimary": tr...Type": "FloatVector", "elementTypeParams": {"dim": "128"}}]}, "indexParams": [{"fieldName": "float_vector", "indexName": "float_vector", "metricType": "IP"}], "params": {"consistencyLevel": "Strong"}}, 
response: {"code":200,"data":{}} (milvus.py:17)
[2024-05-11 11:12:51 - DEBUG - urllib3.connectionpool]: Starting new HTTP connection (1): 10.104.1.35:19530 (connectionpool.py:245)
[2024-05-11 11:12:51 - DEBUG - urllib3.connectionpool]: http://10.104.1.35:19530 "POST /v2/vectordb/collections/describe HTTP/1.1" 200 1087 (connectionpool.py:473)
[2024-05-11 11:12:51 - DEBUG - ci_test]: 
method: post, 
url: http://10.104.1.35:19530/v2/vectordb/collections/describe, 
cost time: 0.1290888786315918, 
header: {'Content-Type': 'application/json', 'Authorization': 'Bearer root:Milvus', 'RequestId': '5a1eb696-0f44-11ef-9c92-acde48001122'}, 
payload: {"collectionName": "test_collection_2024_05_11_11_12_50_440059vwiViLkk"}, 
response: {"code":200,"data":{"aliases":[],"autoId":true,"collectionID":449617770820131178,"collectionName":"test_collection_2024_05_11_11_12_50_440059vwiViLkk","consistencyLevel":"Strong","description":"","ena...e,"type":"FloatVector"}],"indexes":[{"fieldName":"float_vector","indexName":"float_vector","metricType":"IP"}],"load":"LoadStateLoading","partitionsNum":64,"properties":[],"shardsNum":1},"message":""} (milvus.py:17)
[2024-05-11 11:12:51 - INFO - ci_test]: rsp: {'code': 200, 'data': {'aliases': [], 'autoId': True, 'collectionID': 449617770820131178, 'collectionName': 'test_collection_2024_05_11_11_12_50_440059vwiViLkk', 'consistencyLevel': 'Strong', 'description': '', 'enableDynamicField': True, 'fields': [{'autoId': True, 'description': '', 'id': 100, 'name': 'book_id', 'partitionKey': False, 'primaryKey': True, 'type': 'Int64'}, {'autoId': False, 'description': '', 'id': 101, 'name': 'user_id', 'partitionKey': True, 'primaryKey': False, 'type': 'Int64'}, {'autoId': False, 'description': '', 'id': 102, 'name': 'word_count', 'partitionKey': False, 'primaryKey': False, 'type': 'Int64'}, {'autoId': False, 'description': '', 'id': 103, 'name': 'book_describe', 'params': [{'key': 'max_length', 'value': '256'}], 'partitionKey': False, 'primaryKey': False, 'type': 'VarChar'}, {'autoId': False, 'description': '', 'id': 104, 'name': 'float_vector', 'params': [{'key': 'dim', 'value': '128'}], 'partitionKey': False, 'primaryKey': False, 'type': 'FloatVector'}], 'indexes': [{'fieldName': 'float_vector', 'indexName': 'float_vector', 'metricType': 'IP'}], 'load': 'LoadStateLoading', 'partitionsNum': 64, 'properties': [], 'shardsNum': 1}, 'message': ''} (test_vector_operations.py:675)
[2024-05-11 11:12:51 - DEBUG - urllib3.connectionpool]: Starting new HTTP connection (1): 10.104.1.35:19530 (connectionpool.py:245)
[2024-05-11 11:12:52 - DEBUG - urllib3.connectionpool]: http://10.104.1.35:19530 "POST /v2/vectordb/entities/insert HTTP/1.1" 200 None (connectionpool.py:473)
[2024-05-11 11:12:52 - DEBUG - ci_test]: 
method: post, 
url: http://10.104.1.35:19530/v2/vectordb/entities/insert, 
cost time: 0.6702909469604492, 
header: {'Content-Type': 'application/json', 'Authorization': 'Bearer root:Milvus', 'Accept-Type-Allow-Int64': 'true', 'RequestId': '5a37237a-0f44-11ef-9c92-acde48001122'}, 
payload: {"collectionName": "test_collection_2024_05_11_11_12_50_440059vwiViLkk", "data": [{"user_id": 0, "word_count": 0, "book_describe": "book_0", "float_vector": [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, ..., 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], "dynamic_field_2999": 2999}]}, 
response: {"code":200,"data":{"insertCount":3000,"insertIds":[449617770818713663,449617770818713664,449617770818713665,449617770818713666,449617770818713667,449617770818713668,449617770818713669,449617770818713...8716652,449617770818716653,449617770818716654,449617770818716655,449617770818716656,449617770818716657,449617770818716658,449617770818716659,449617770818716660,449617770818716661,449617770818716662]}} (milvus.py:17)
[2024-05-11 11:12:53 - DEBUG - urllib3.connectionpool]: Starting new HTTP connection (1): 10.104.1.35:19530 (connectionpool.py:245)
[2024-05-11 11:12:56 - DEBUG - urllib3.connectionpool]: http://10.104.1.35:19530 "POST /v2/vectordb/entities/search HTTP/1.1" 200 None (connectionpool.py:473)
[2024-05-11 11:12:56 - DEBUG - ci_test]: 
method: post, 
url: http://10.104.1.35:19530/v2/vectordb/entities/search, 
cost time: 3.0752687454223633, 
header: {'Content-Type': 'application/json', 'Authorization': 'Bearer root:Milvus', 'Accept-Type-Allow-Int64': 'true', 'RequestId': '5b40110a-0f44-11ef-9c92-acde48001122'}, 
payload: {"collectionName": "test_collection_2024_05_11_11_12_50_440059vwiViLkk", "data": [[0.07760931870102615, 0.11648295978031525, 0.06759402962605651, 0.07627516229539759, 0.09439156081631465, 0.0089967893...473742655506]], "filter": "word_count > 100", "groupingField": "user_id", "outputFields": ["*"], "searchParams": {"metricType": "IP", "params": {"radius": "0.1", "range_filter": "0.8"}}, "limit": 100}, 
response: {"code":200,"data":[{"book_describe":"book_2460","book_id":449617770818716123,"distance":0,"dynamic_field_2460":2460,"float_vector":[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0...0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],"id":449617770818716662,"user_id":99,"word_count":2999}]} (milvus.py:17)
PASSED

[xiaofan-luan]

The entities in question have zero vectors and that seem to trigger the nan. It should be 0 instead of nan.

how to represent nan in restful?

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions. Rotten issues close after 30d of inactivity. Reopen the issue with /reopen.