ImportError: cannot import name 'bool' from 'numpy'

How to fix ImportError: cannot import name 'bool' from 'numpy'

Stop importing the alias. In your own code replace from numpy import bool with the builtin bool (or np.bool_ if you need the numpy scalar type). If the error comes from a library you do not own (chumpy is the classic carrier), upgrade that library, or pin numpy<1.24 as a stopgap. numpy 1.24 removed the old aliases np.bool, np.int, np.float and friends for good.

The exact error

ImportError: cannot import name 'bool' from 'numpy'

Fix it

1. Find the bad import: grep -rn "from numpy import" --include=*.py . and look for bool, int, float, complex, object, unicode, str. Expected: you see the exact line raising the error

2. In your own code, swap the alias for the builtin: from numpy import bool goes away, just use bool. Only keep numpy's np.bool_ if you truly need the numpy scalar. Expected: import succeeds on numpy>=1.24

3. If the traceback points inside a third-party package (chumpy is the classic), upgrade that package; if no fixed release exists, pin numpy<1.24 in your requirements. Expected: pip shows numpy 1.23.x and the import works

When this applies

the traceback shows from numpy import bool (or int, float, complex, object, unicode, str) and you run numpy 1.24 or newer

When this does NOT apply

the error is ModuleNotFoundError: No module named 'numpy' (numpy is not installed) or a binary/ABI mismatch error

Versions

numpy >= 1.24, Python 3.8+

Why it happens

numpy deprecated these aliases in 1.20 and removed them in 1.24. Old code and old libraries still import them.

Edge cases

np.bool_ still exists and is fine to use. Pinning numpy<1.24 is a band-aid; the real fix is updating the importing code.

Asked about: ImportError: cannot import name 'bool' from 'numpy' (gh:jxyjxyw/stereo-inertial-poser (README known-issues)). The verified fix: Stop importing the alias. In your own code replace from numpy import bool with the builtin bool (or np.bool_ if you need the numpy scalar type). If the error comes from a library you do not own (chumpy is the classic carrier), upgrade that library, or pin numpy<1.24 as a stopgap. numpy 1.24 removed the old aliases np.bool, np.int, np.float and friends for good. Full thread: https://github.com/jxyjxyw/stereo-inertial-poser/blob/HEAD/README.md