*   TensorFlow pip packages are now built with CUDA11.2 and cuDNN 8.1.0

## Breaking Changes

*   The `TF_CPP_MIN_VLOG_LEVEL` environment variable has been renamed to
    `TF_CPP_MAX_VLOG_LEVEL` which correctly describes its effect.

## Bug Fixes and Other Changes

*   `tf.keras`:

    *   Preprocessing layers API consistency changes:
        *   `StringLookup` added `output_mode`, `sparse`, and

- Fix kubectl describe CSINode nil pointer error ([#89646](https://github.com/kubernetes/kubernetes/pull/89646), [@zhouya0](https://github.com/zhouya0)) [SIG CLI]
- Fix kubectl describe node for users not having access to lease information. ([#90469](https://github.com/kubernetes/kubernetes/pull/90469), [@uthark](https://github.com/uthark)) [SIG CLI]

* In `encoders.jsonable_encoder` remove parameter `sqlalchemy_safe`.
    * It was an early hack to allow returning SQLAlchemy models, but it was never documented, and the recommended way is using Pydantic's `orm_mode` as described in the tutorial: [SQL (Relational) Databases](https://fastapi.tiangolo.com/tutorial/sql-databases/).
    * PR [#1864](https://github.com/tiangolo/fastapi/pull/1864).

### Docs

b.limbs[:n] switch n { default: // Attempt to use a stack-allocated backing array. T := make([]uint, 0, preallocLimbs*2) if cap(T) < n*2 { T = make([]uint, 0, n*2) } T = T[:n*2] // This loop implements Word-by-Word Montgomery Multiplication, as // described in Algorithm 4 (Fig. 3) of "Efficient Software // Implementations of Modular Exponentiation" by Shay Gueron // [https://eprint.iacr.org/2011/239.pdf]. var c uint for i := 0; i < n; i++ { _ = T[n+i] // bounds check elimination hint // Step 1 (T =...