Hi, did anyone have this error while trying to run

containerize

?

0.166 Installing pip packages from 'requirements.lock.txt'..
0.167 + pip_install -r /home/bentoml/bento/env/python/requirements.lock.txt
0.167 + command -v uv
0.167 + uv pip install -r /home/bentoml/bento/env/python/requirements.lock.txt
0.209   × No solution found when resolving dependencies:
0.209   ╰─▶ Because the current Python version (3.9.20) does not satisfy
0.209       Python>=3.10 and numpy==2.1.1 depends on Python>=3.10, we can conclude
0.209       that numpy==2.1.1 cannot be used.
0.209       And because you require numpy==2.1.1, we can conclude that your
0.209       requirements are unsatisfiable.

I tried running with an empty requirements file or a single requirement that is not numpy and still got the error. this is what i added to the bentofile:

python:
  requirements_txt: "./requirements.txt"
docker:
  distro: debian
  python_version: "3.9"

and when deleting the requirements_txt from the bentofile everything worked fine

We are using python 3.9 and have pinned the following in our requirements.txt:

numpy==1.24.4   # numpy 2.0 breaks backward compatibility on existing libraries
scipy==1.13.1   # scipy 1.14.0 requires python 3.10+
dask==2024.6.2  # dask==2024.7.0 removed pandas 2 support. that breaks bentoml build.
pydantic==2.8.2

Thanks, that indeed solves the problem, I just don't understand where this requirement is coming from

They are transient dependencies that bentoml dependent on but were upgraded independently and broke bentoml. It is not easy for an open-source library like bentoml to pin specific versions while preserving certain backward compatibility to previous bentoml versions. We had to find out in the hard way.

thanks for the help!

CC: @Frost Ming

build bento on python 3.9 as well

changing

lock_packages

to

false

worked, thank you very much!