9cfe078a - [#3572](https://github.com/google/flax/pull/3572) was rolled back because of internal test breakages. This PR re-adds the attention changes again and fixes internal tests
[#3572](https://github.com/google/flax/pull/3572) was rolled back because of internal test breakages. This PR re-adds the attention changes again and fixes internal tests
PiperOrigin-RevId: 597881774