# What changed in tensorflow

> 13 changes in tensorflow 2 from its official release notes, upgrade guides and advisories: 2 breaking, 6 new APIs and options. Each one cited. Sources checked 2026-10-18. HTML: https://unstale.sh/pypi/tensorflow/

To see which of these apply to a repository, run `npx @unstale/cli init` in it.

## tensorflow 2

### Breaking changes

- Support for Python 3.9 has been removed, so developers must use a supported Python version. (2.21.0) — [release notes](https://github.com/tensorflow/tensorflow/releases/tag/v2.21.0)
- The TensorBoard dependency has been removed from TensorFlow. (2.21.0) — [release notes](https://github.com/tensorflow/tensorflow/releases/tag/v2.21.0)

### New APIs and options

- The NOT_EQUAL operator adds int16x8 support in tf.lite. (2.21.0) — [release notes](https://github.com/tensorflow/tensorflow/releases/tag/v2.21.0)
- The SQRT operator adds int8 support in tf.lite. (2.21.0) — [release notes](https://github.com/tensorflow/tensorflow/releases/tag/v2.21.0)
- The SQRT operator adds int16x8 support in tf.lite. (2.21.0) — [release notes](https://github.com/tensorflow/tensorflow/releases/tag/v2.21.0)
- tf.data adds NoneTensorSpec to the public API so None values in element_spec can be identified with isinstance. (2.21.0) — [release notes](https://github.com/tensorflow/tensorflow/releases/tag/v2.21.0)
- tfl.cast adds support for int2 and int4. (2.21.0) — [release notes](https://github.com/tensorflow/tensorflow/releases/tag/v2.21.0)
- tf.lite adds support for the uint4 type. (2.21.0) — [release notes](https://github.com/tensorflow/tensorflow/releases/tag/v2.21.0)

### Behavior and default changes

- The EQUAL operator adds int16x8 support in tf.lite. (2.21.0) — [release notes](https://github.com/tensorflow/tensorflow/releases/tag/v2.21.0)
- tf.image.decode_image adds JPEG XL support. (2.21.0) — [release notes](https://github.com/tensorflow/tensorflow/releases/tag/v2.21.0)
- tfl.fully_connected adds support for SRQ int2. (2.21.0) — [release notes](https://github.com/tensorflow/tensorflow/releases/tag/v2.21.0)

### Fixes and other changes

- tf.lite adds support for the int2 type. (2.21.0) — [release notes](https://github.com/tensorflow/tensorflow/releases/tag/v2.21.0)
- tfl.slice adds support for int4. (2.21.0) — [release notes](https://github.com/tensorflow/tensorflow/releases/tag/v2.21.0)