X Tutup
Skip to content

[Bug]: semantic_memory.py _background_ingestion_task crashes frequently when running amemgym benchmark #1144

@tomw-mv

Description

@tomw-mv

Describe the bug

I am running amemgym benchmark using memmachine semantic memory:

https://github.com/AGI-Eval-Official/amemgym

While running eval, observe large number of semantic memory exceptions:

tomw@mac-work Desktop % grep '^Traceback' memmachine.stderr.txt|wc -l
903

it says task crashed:

 2026-02-26 18:34:51,156 [ERROR] memmachine.semantic_memory.semantic_memory - background task crashed, restarting

I suspect half of the add_memory operations are not completed due to the large amount of crashes.

Steps to reproduce

Info on running the benchmark is here:

https://memverge.slack.com/archives/C0A0Z69UC8L/p1771976896651179

Expected behavior

Add memory succeeds

Environment

Build is 2/9 main aaea7b7

Additional context

memmachine.stderr.txt

Metadata

Metadata

Assignees

Labels

No labels
No labels

Type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions

    X Tutup