Segmentation fault when using multiprocessing

[arshsingh]

Example:

from multiprocessing import Process
from baml_client import b


def run():
    b.ExtractResume("test")


if __name__ == "__main__":
    current = Process(target=run)
    current.start()
    current.join()

This exits with a seg fault.

lazy import works though

from multiprocessing import Process

def run():
    from baml_client import b
    b.ExtractResume("test")


if __name__ == "__main__":
    current = Process(target=run)
    current.start()
    current.join()

I can reproduce this with 0.82.0 on the following platforms:

• python:3.12 arm64 docker image
• python:3.10 arm64 docker image
• python:3.12 amd64 docker image
• amazon linux 2023 with python 3.9

Seems to be working without any issues on my M2 mac though (python 3.10).

[linear[bot]]

BAML-317 Segmentation fault when using multiprocessing

[mentatbot[bot]]

If you would like me to solve this issue, either tag me in a comment or check this box:

• [ ] Solve Issue

_{You can disable automatic comments on my settings page}

[w0r1dhe110]

Has this issue been resolved? I'm encountering the same error.

• baml 0.89.0
• python:3.12 amd64 docker image

[aaronvg]

follow this for the resolution: https://github.com/BoundaryML/baml/issues/1821 We are looking into it now

[ba11b0y]

#1990 fixes this