Segmentation fault when calling kb.get_vector

[Joozty]

How to reproduce the behaviour

When I run this example I get a segmentation fault. I am not sure how to get a stacktrace. I tried to set this env variable PYTHONFAULTHANDLER=1 but I get only this:

Fatal Python error: Segmentation fault

Current thread 0x000000010099fd40 (most recent call first):
  File "/Users/joozty/Documents/kb_test.py", line 24 in <module>

Process finished with exit code 139 (interrupted by signal 11: SIGSEGV)

I should note that I am using a new MacBook with an M1 chip so it can be something platform specific. Also, it happens only when I serialize KB which was generated with set_entities. When I add entity one by one it works just fine.

Code to reproduce:

from spacy.kb import KnowledgeBase
import spacy

# works fine
# nlp = spacy.load("en_core_web_md")
# kb = KnowledgeBase(vocab=nlp.vocab, entity_vector_length=300)
# kb.add_entity("1", 1, nlp("test").vector)
# kb.to_disk("test")
#
# kb1 = KnowledgeBase(vocab=nlp.vocab, entity_vector_length=300)
# kb1.from_disk("test")
# kb1.get_vector("1")



# fails with Segmentation fault
nlp = spacy.load("en_core_web_md")
kb = KnowledgeBase(vocab=nlp.vocab, entity_vector_length=300)
kb.set_entities(["1"], [1], [nlp("test").vector])
kb.to_disk("test")

kb1 = KnowledgeBase(vocab=nlp.vocab, entity_vector_length=300)
kb1.from_disk("test")
kb1.get_vector("1")

Your Environment

Info about spaCy

• spaCy version: 3.1.2
• Platform: macOS-11.5.2-arm64-arm-64bit
• Python version: 3.9.6
• Pipelines: xx_ent_wiki_sm (3.1.0), en_core_web_sm (3.1.0), en_core_web_md (3.1.0)

[adrianeboyd]

Thanks for the report! I can reproduce this error and it doesn't look platform-specific. We will look into it!

[WMRamadan]

I can confirm this problem can be reproduced on the following:
OS: Ubuntu Linux 21.04
Kernel: 5.11.0-34-generic
Python: 3.9.5
Pip: 20.3.4
spaCy: 3.1.3

A trace revealed the problem is related to the following method https://github.com/explosion/spaCy/blob/master/spacy/kb.pyx#L307

I believe specifically to https://github.com/explosion/spaCy/blob/master/spacy/kb.pyx#L315
where self._vectors_table[] appears to be the problem.

Which is initialized here https://github.com/explosion/spaCy/blob/master/spacy/kb.pyx#L99

Where float_matrix is being declared from spacy/kb.pyd here https://github.com/explosion/spaCy/blob/master/spacy/kb.pxd#L16

[svlandeg]

Hi all, thanks for your patience! An explanation and solution for this bug is provided in #9463.

[github-actions]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.