[Mlir-commits] [mlir] [mlir][python] Raise maximum allowed version (PR #114050)

[Scott Todd]

================
@@ -1,4 +1,4 @@
-numpy>=1.19.5, <=1.26
-pybind11>=2.9.0, <=2.10.3
+numpy>=1.19.5, <=2.1.2
----------------
ScottTodd wrote:

How much can we trust the CI checks for this version upgrade? Moving from numpy 1 --> 2 caused (is still causing) quite a bit of ecosystem churn (https://numpy.org/doc/stable/numpy_2_0_migration_guide.html).

Is this a "merge and see if anyone complains" scenario? Or should some other tests/experiments be run or the change be announced on Discourse?

https://github.com/llvm/llvm-project/pull/114050