deeplake 3.9.52 incompatible with NumPy 2.x (NEP 50): TypeError in get_incompatible_dtype

[LuigiGonnella]

Bug Report

Environment

• deeplake==3.9.52
• numpy>=2.0 (tested on 2.4.2)

Description

NumPy 2.x (NEP 50) removed support for passing raw Python scalars (int, float, bool) as the first argument to np.can_cast. This causes a TypeError when DeepLake attempts to validate scalar samples in get_incompatible_dtype inside deeplake/util/casting.py.

Error

TypeError: can_cast() does not support Python ints, floats, and complex because the result used to depend on the value.

The offending line in deeplake/util/casting.py:

if np.can_cast(samples, dtype)  # samples is a raw Python float/int

Fix

A fix is available in this PR:
#3143

References

• NumPy 2.0 migration guide
• NEP 50 — Promotion rules for Python scalars

[ferhimedamine]

The NumPy 2.x incompatibility (NEP 50 type promotion changes) is affecting a lot of libraries — NEP 50 changed how Python scalars interact with NumPy arrays, which breaks code that relied on implicit broadcasting behavior.

The fix is usually to add explicit dtype casts wherever the logic runs, replacing any calls that mix Python scalars and NumPy arrays with explicit numpy dtype arguments.

For teams using DeepLake as a vector store for agent memory who hit this on upgrade, pinning NumPy to is a quick workaround while the proper fix lands. If you're also looking for a self-hosted memory alternative with a simple REST API that doesn't have this dependency, dakera-deploy might be worth evaluating.