March 2026 monthly performance summary for pymc-devs/pytensor: Focused on improving robustness and reliability of tensor operations by expanding tests for the Kron function to cover both static and dynamic shapes. The change reduces shape-related regressions and increases confidence for downstream users in PyMC workflows.

February 2026 monthly summary for pymc-devs/pytensor: Focused on reliability in static shape inference for linear algebra operations. Delivered a Kron shape inference bug fix and added a regression test to prevent future regressions. This work reduces runtime shape errors in downstream PyMC models and strengthens maintainability of the linear algebra module.

January 2026 focused on accelerating developer productivity and code quality across core ML stacks (keras-team/keras and google/flax). Delivered documentation improvements, API configurability enhancements, and a new attention capability, translating to faster onboarding, easier feature usage, and more robust model architectures. The work reduces ambiguity for users and empowers teams to experiment with configurable layers and attention mechanisms with greater confidence.