I’m getting this problem when running dbt with system deets:

Running with dbt=1.5.9
Registered adapter: fal=1.5.4
Registered adapter: postgres=1.5.9

on a project with fal-project.yml:

environments:
  - name: ml
    type: venv
    requirements:
      - scipy
      - pandas
      - numpy
      - statsmodels
      - catboost

The obvious fix per the accepted answer to the issue linked above, and many other answers, is to specify numpy==1.26.4. This does not fix the problem. I tried also pinning pandas to the prevailing version from back in Jan 2024 (2.2.0) when my repo definitely worked, but this doesn’t help either.

Even trying to run just dbt --version gives the error, with stacktrace ending at

    from pandas._libs.interval import Interval
  File “pandas/_libs/interval.pyx”, line 1, in init pandas._libs.interval
ValueError: numpy.dtype size changed, may indicate binary incompatibility. Expected 96 from C header, got 88 from PyObject

It behaves as if fal is ignoring the attempt to pin the version. I infer from this that the syntax is correct, however.

How can I force the numpy-downgrading fix that appears to be widely known?