*   Provided a new `experimental_skip_saver` argument which, if specified, will suppress the addition of `SavedModel`-native save and restore ops to the `SavedModel`, for cases where users already build custom save/restore ops and checkpoint formats for the model being saved, and the creation of the SavedModel-native save/restore ops simply cause longer model serialization times.

* Add ops to `tensorflow.raw_ops` that were missing.

* `tf.CheckpointOptions`

* ✨ Add support for Pydantic models in `Form` parameters. PR [#12127](https://github.com/fastapi/fastapi/pull/12127) by [@tiangolo](https://github.com/tiangolo). Reverted by PR [#12128](https://github.com/fastapi/fastapi/pull/12128) to make a checkpoint release with only refactors. Restored by PR [#12129](https://github.com/fastapi/fastapi/pull/12129).

* * T1 = Z1*Z1 * T2 = Z2*Z2 * U1 = X1*T2 * H = X2*T1 * H = H-U1 * Z3 = Z1*Z2 * Z3 = Z3*H << store-out Z3 result reg.. could override Z1, if slices have same backing array * * S1 = Z2*T2 * S1 = Y1*S1 * R = Z1*T1 * R = Y2*R * R = R-S1 * * T1 = H*H * T2 = H*T1 * U1 = U1*T1 * * X3 = R*R * X3 = X3-T2 * T1 = 2*U1 * X3 = X3-T1 << store-out X3 result reg * * T2 = S1*T2 * Y3 = U1-X3 * Y3 = R*Y3 * Y3 = Y3-T2 << store-out Y3 result reg // X=Z1; Y=Z1; MUL; T- // T1 = Z1*Z1 // X- ; Y=T ; MUL; R=T // R = Z1*T1...

* * T1 = Z1*Z1 * T2 = Z2*Z2 * U1 = X1*T2 * H = X2*T1 * H = H-U1 * Z3 = Z1*Z2 * Z3 = Z3*H << store-out Z3 result reg.. could override Z1, if slices have same backing array * * S1 = Z2*T2 * S1 = Y1*S1 * R = Z1*T1 * R = Y2*R * R = R-S1 * * T1 = H*H * T2 = H*T1 * U1 = U1*T1 * * X3 = R*R * X3 = X3-T2 * T1 = 2*U1 * X3 = X3-T1 << store-out X3 result reg * * T2 = S1*T2 * Y3 = U1-X3 * Y3 = R*Y3 * Y3 = Y3-T2 << store-out Y3 result reg // X=Z1; Y=Z1; MUL; T- // T1 = Z1*Z1 // X- ; Y=T ; MUL; R=T // R = Z1*T1...