# tensorflow changelog

> An Open Source Machine Learning Framework for Everyone

- Vendor: tensorflow
- Category: AI
- Official site: https://tensorflow.org/
- Tracked by: What's New (https://whatsnew.fyi/product/tensorflow)
- Harvested from: GitHub (tensorflow/tensorflow)
- Entries below: 10 (newest first)

What's New is an index, not a publisher: every entry below links to the vendor's own release notes, which are the authoritative source. Entries are labelled where they are hand-curated sample data, pre-releases, or drawn from a secondary source such as a developer blog.

## Releases

### v2.21.0 — TensorFlow 2.21.0

- Date: 2026-03-06
- Version: v2.21.0
- Original notes: https://github.com/tensorflow/tensorflow/releases/tag/v2.21.0
- Permalink: https://whatsnew.fyi/product/tensorflow/releases/v2.21.0

- **removed** — Support for Python 3.9 has been removed starting with TF 2.21
- **removed** — The TensorBoard (TB) dependency has been removed starting with TF 2.21
- **added** — Add int8 and int16x8 support for SQRT operator in tf.lite
- **added** — Add int16x8 support for EQUAL and NOT_EQUAL operators in tf.lite
- **added** — Add support for int2 type in tf.lite
- **added** — Add support for int2 and int4 in tfl.cast
- **added** — Add support for SRQ int2 in tfl.fully_connected
- **added** — Add support for int4 in tfl.slice
- **added** — Add support for uint4 type in tf.lite
- **added** — Add JPEG XL support in tf.image.decode_image
- **added** — Add NoneTensorSpec to the public API so that Nones in element_spec can be identified via isinstance

#### Release 2.21.0

##### TensorFlow

###### Breaking Changes

* Support for Python 3.9 has been removed starting with TF 2.21.
* The TensorBoard (TB) dependency has been removed starting with TF 2.21.

###### Major Features and Improvements

* `tf.lite`
    * Adds int8 and int16x8 support for SQRT operator.
    * Adds int16x8 support for EQUAL and NOT_EQUAL operators.
    * Adds support for int2 type.
    * Adds support for int2/int4 in tfl.cast .
    * Adds support for SRQ int2 in tfl.fully_connected.
    * Adds support for int4 in tfl.slice.
    * Adds support for uint4 type.

*  `tf.image`
    * Adds JPEG XL support in decode_image.

###### Bug Fixes and Other Changes

* `tf.data`
    * Adds `NoneTensorSpec` to the public API so that `None`s in `element_spec`
      can be identified via `isinstance(..., tf.NoneTensorSpec)`.

##### Thanks to our Contributors

This release contains contributions from many people at Google, as well as:

Aaraviitkgp, Abhijeet, Abhinav Gunjal, Abhishek, Adam Paszke, Aditya Gupta, Aditya Jha, Aditya Sharma, Adrian Kuegel, Aiden Grossman, Akarsh, Akhil Goel, Alan Kelly, Aleksa Arsic, Aleksei, Aleksei Nurmukhametov, Alex, Alexander Belyaev, Alexander Grund, Alexander Lyashuk, Alexander Shaposhnikov, Alex Pivovarov, Aliia Khasanova, Alina Sbirlea, Allan Renucci, Amelia Thurdekoos, Amit Sabne, Andrei Ivanov, Andrew Dame, Andrey Portnoy, Anish Nair, Anlun Xu, Antonio Sanchez, anuj chincholikar, Anuj Chincholikar, Aravindh Balaji, aravindhbalaji1985, Arian Arfaian, Armin Felder, Artem Belevich, Ashish Rao, Ashitesh Singh, A. Unique TensorFlower, Bart Chrzaszcz, benediktjohannes, Benjamin Chetioui, Benjamin Kramer, Berkin Ilbeyi, Bhatu, Bhavani Subramanian, Bhupendra Dubey, Bill Varcho, Bixia Zheng, Blake Hechtman, Bodhi Silberling, BruceXinXin, Bryan Massoth, Buddh Prakash, Byungchul Kim, Ce Zheng, Changhui Lin, Chao, Charles Alaras, Chase Riley Roberts, Chenhao Jiang, Chris Ashton, Chris Jones, Chris Kennelly, Christian Sigg, Chuan He, Chunlei Niu, Chun-nien Chan, Chunyu Jin, Clive Verghese, Cong Liu, Corentin Kerisit, Daniel Chen, Daniel Kuts, Daniel Ng, Daniel Sosa, Daniel Suo, Danila Malyutin, David Dunleavy, David Majnemer, David Pizzuto, Deepika Rajani, deeptanshusekhri, dependabot[bot], Deqiang Chen, Derek Murray, Dillon Sharlet, Dimitar (Mitko) Asenov, Dimitris Vardoulakis, Dirk Hornung, DottsGit, Dragan Mladjenovic, Eetu Sjöblom, Elen Kalda, Emilio Cota, Emily Fertig, Eugene Zhulenev, Eusebio Durán Montaña, Evan Brown, Ezekiel Calubaquib, Faijul Amin, Felix Wang, Fengwu Yao, Fergus Henderson, Frederic Rechtenstein, Frederik Gossen, Gabriel Gerlero, Gagan Nagaraj, gaikwadrahul8, garry00107, gaurides, George Pawelczak, Georg Stefan Schmid, gns, Goran Flegar, Graham, Grant Jensen, Greg Olechwierowicz, Gregory Pataky, Grzegorz Gawryał, Gunhyun Park, guozhong.zhuang, Haibo Huang, Hana Joo, Hariprasad Ravishankar, Harsha H S, Harshit Monish, Henning Becker, Hittanshu, Hoeseong (Hayden) Kim, Hugo Mano, Hyeontaek Lim, Ibrahim Umit Akgun, ILCSFNO, Ilia Sergachev, Ilya Tikhonovskiy, Iman Hosseini, Ionel Gog, Isha Arkatkar, isharif168, Ivo Ristovski List, Jacques Pienaar, Jae H. Yoo, Jaeyoon Jung, Jake Harmon, James Hilliard, jameslovespancakes, James Spooner, Jane Liu, Jaroslav Sevcik, Jeff Parker, Jeffrey A. Dean, Jeremy Meredith, Jialei Chen, Jian Cai, Jian Li, Jie Luo, Jim Lin, Jing Pu, Jinliang Wei, Jiya Zhang, Joel Wee, Johannes Buchner, Johannes Reifferscheid, Johnny, Jorge Gorbe Moya, Joshua Lang, Joshua Wang, Joss Briody, jparkerh, Juanli Shen, Juhyun Lee, Jun Jiang, Junwhan Ahn, Kadir Barut, Kanglan Tang, Kanish Anand, Kanvi Khanna, Karlo Basioli, Ken Franko, Kevin Chen, Kevin Gleason, Kingston Mandisodza, Koki Ibukuro, Kostiantyn Liepieshov, Krishna Haridasan, Krishna Somani, Krzysztof Kosiński, Kuy Mainwaring, lambert, Larry Lansing, Lin Chai, Lord ε Rebel, Luke Baumann, Luke Hutton, madhavmadupu, Majid Dadashi, Mani Ananth, Manjunath Gaonkar, Marcello Maggi

_[Truncated at 4000 characters — full notes: https://github.com/tensorflow/tensorflow/releases/tag/v2.21.0]_

### v2.21.0-rc1 — TensorFlow 2.21.0-rc1

- Date: 2026-03-02
- Version: v2.21.0-rc1
- Original notes: https://github.com/tensorflow/tensorflow/releases/tag/v2.21.0-rc1
- Permalink: https://whatsnew.fyi/product/tensorflow/releases/v2.21.0-rc1
- Labels: Pre-release

- **removed** — Support for Python 3.9 has been removed
- **added** — Add int8 and int16x8 support for SQRT operator in tf.lite
- **added** — Add int16x8 support for EQUAL and NOT_EQUAL operators in tf.lite
- **added** — Add support for int2 type in tf.lite
- **added** — Add support for int2/int4 in tfl.cast
- **added** — Add support for SRQ int2 in tfl.fully_connected
- **added** — Add support for int4 in tfl.slice
- **added** — Add support for uint4 type in tf.lite
- **added** — Add JPEG XL support in tf.image.decode_image
- **added** — Add NoneTensorSpec to the public API so that Nones in element_spec can be identified via isinstance

#### Release 2.21.0

##### TensorFlow

###### Breaking Changes

* Support for Python 3.9 has been removed starting with TF 2.21.

###### Major Features and Improvements

* `tf.lite`
    * Adds int8 and int16x8 support for SQRT operator.
    * Adds int16x8 support for EQUAL and NOT_EQUAL operators.
    * Adds support for int2 type.
    * Adds support for int2/int4 in tfl.cast .
    * Adds support for SRQ int2 in tfl.fully_connected.
    * Adds support for int4 in tfl.slice.
    * Adds support for uint4 type.

*  `tf.image`
    * Adds JPEG XL support in decode_image.

###### Bug Fixes and Other Changes

* `tf.data`
    * Adds `NoneTensorSpec` to the public API so that `None`s in `element_spec`
      can be identified via `isinstance(..., tf.NoneTensorSpec)`.

##### Thanks to our Contributors

This release contains contributions from many people at Google, as well as:

Aaraviitkgp, Abhijeet, Abhinav Gunjal, Abhishek, Adam Paszke, Aditya Gupta, Aditya Jha, Aditya Sharma, Adrian Kuegel, Aiden Grossman, Akarsh, Akhil Goel, Alan Kelly, Aleksa Arsic, Aleksei, Aleksei Nurmukhametov, Alex, Alexander Belyaev, Alexander Grund, Alexander Lyashuk, Alexander Shaposhnikov, Alex Pivovarov, Aliia Khasanova, Alina Sbirlea, Allan Renucci, Amelia Thurdekoos, Amit Sabne, Andrei Ivanov, Andrew Dame, Andrey Portnoy, Anish Nair, Anlun Xu, Antonio Sanchez, anuj chincholikar, Anuj Chincholikar, Aravindh Balaji, aravindhbalaji1985, Arian Arfaian, Armin Felder, Artem Belevich, Ashish Rao, Ashitesh Singh, A. Unique TensorFlower, Bart Chrzaszcz, benediktjohannes, Benjamin Chetioui, Benjamin Kramer, Berkin Ilbeyi, Bhatu, Bhavani Subramanian, Bhupendra Dubey, Bill Varcho, Bixia Zheng, Blake Hechtman, Bodhi Silberling, BruceXinXin, Bryan Massoth, Buddh Prakash, Byungchul Kim, Ce Zheng, Changhui Lin, Chao, Charles Alaras, Chase Riley Roberts, Chenhao Jiang, Chris Ashton, Chris Jones, Chris Kennelly, Christian Sigg, Chuan He, Chunlei Niu, Chun-nien Chan, Chunyu Jin, Clive Verghese, Cong Liu, Corentin Kerisit, Daniel Chen, Daniel Kuts, Daniel Ng, Daniel Sosa, Daniel Suo, Danila Malyutin, David Dunleavy, David Majnemer, David Pizzuto, Deepika Rajani, deeptanshusekhri, dependabot[bot], Deqiang Chen, Derek Murray, Dillon Sharlet, Dimitar (Mitko) Asenov, Dimitris Vardoulakis, Dirk Hornung, DottsGit, Dragan Mladjenovic, Eetu Sjöblom, Elen Kalda, Emilio Cota, Emily Fertig, Eugene Zhulenev, Eusebio Durán Montaña, Evan Brown, Ezekiel Calubaquib, Faijul Amin, Felix Wang, Fengwu Yao, Fergus Henderson, Frederic Rechtenstein, Frederik Gossen, Gabriel Gerlero, Gagan Nagaraj, gaikwadrahul8, garry00107, gaurides, George Pawelczak, Georg Stefan Schmid, gns, Goran Flegar, Graham, Grant Jensen, Greg Olechwierowicz, Gregory Pataky, Grzegorz Gawryał, Gunhyun Park, guozhong.zhuang, Haibo Huang, Hana Joo, Hariprasad Ravishankar, Harsha H S, Harshit Monish, Henning Becker, Hittanshu, Hoeseong (Hayden) Kim, Hugo Mano, Hyeontaek Lim, Ibrahim Umit Akgun, ILCSFNO, Ilia Sergachev, Ilya Tikhonovskiy, Iman Hosseini, Ionel Gog, Isha Arkatkar, isharif168, Ivo Ristovski List, Jacques Pienaar, Jae H. Yoo, Jaeyoon Jung, Jake Harmon, James Hilliard, jameslovespancakes, James Spooner, Jane Liu, Jaroslav Sevcik, Jeff Parker, Jeffrey A. Dean, Jeremy Meredith, Jialei Chen, Jian Cai, Jian Li, Jie Luo, Jim Lin, Jing Pu, Jinliang Wei, Jiya Zhang, Joel Wee, Johannes Buchner, Johannes Reifferscheid, Johnny, Jorge Gorbe Moya, Joshua Lang, Joshua Wang, Joss Briody, jparkerh, Juanli Shen, Juhyun Lee, Jun Jiang, Junwhan Ahn, Kadir Barut, Kanglan Tang, Kanish Anand, Kanvi Khanna, Karlo Basioli, Ken Franko, Kevin Chen, Kevin Gleason, Kingston Mandisodza, Koki Ibukuro, Kostiantyn Liepieshov, Krishna Haridasan, Krishna Somani, Krzysztof Kosiński, Kuy Mainwaring, lambert, Larry Lansing, Lin Chai, Lord ε Rebel, Luke Baumann, Luke Hutton, madhavmadupu, Majid Dadashi, Mani Ananth, Manjunath Gaonkar, Marcello Maggioni, Marcin Radomski, Maria Lyubimtseva, Marissa Ikonomidis, Mark Daoust, M

_[Truncated at 4000 characters — full notes: https://github.com/tensorflow/tensorflow/releases/tag/v2.21.0-rc1]_

### v2.21.0-rc0 — TensorFlow 2.21.0-rc0

- Date: 2026-02-09
- Version: v2.21.0-rc0
- Original notes: https://github.com/tensorflow/tensorflow/releases/tag/v2.21.0-rc0
- Permalink: https://whatsnew.fyi/product/tensorflow/releases/v2.21.0-rc0
- Labels: Pre-release

- **removed** — Support for Python 3.9 has been removed
- **added** — Add int8 and int16x8 support for SQRT operator in tf.lite
- **added** — Add int16x8 support for EQUAL and NOT_EQUAL operators in tf.lite
- **added** — Add support for int2 type in tf.lite
- **added** — Add support for int2 and int4 in tfl.cast
- **added** — Add support for SRQ int2 in tfl.fully_connected
- **added** — Add support for int4 in tfl.slice
- **added** — Add support for uint4 type in tf.lite
- **added** — Add JPEG XL support in decode_image for tf.image
- **added** — Add NoneTensorSpec to the public API in tf.data so that Nones in element_spec can be identified via isinstance

#### Release 2.21.0

##### TensorFlow

###### Breaking Changes

* Support for Python 3.9 has been removed starting with TF 2.21.

###### Major Features and Improvements

* `tf.lite`
    * Adds int8 and int16x8 support for SQRT operator.
    * Adds int16x8 support for EQUAL and NOT_EQUAL operators.
    * Adds support for int2 type.
    * Adds support for int2/int4 in tfl.cast .
    * Adds support for SRQ int2 in tfl.fully_connected.
    * Adds support for int4 in tfl.slice.
    * Adds support for uint4 type.

*  `tf.image`
    * Adds JPEG XL support in decode_image.

###### Bug Fixes and Other Changes

* `tf.data`
    * Adds `NoneTensorSpec` to the public API so that `None`s in `element_spec` can be identified via `isinstance(..., tf.NoneTensorSpec)`.

##### Thanks to our Contributors

This release contains contributions from many people at Google, as well as:

Aaraviitkgp, Abhijeet, Abhinav Gunjal, Abhishek, Adam Paszke, Aditya Gupta, Aditya Jha, Aditya Sharma, Adrian Kuegel, Aiden Grossman, Akarsh, Akhil Goel, Alan Kelly, Aleksa Arsic, Aleksei, Aleksei Nurmukhametov, Alex, Alexander Belyaev, Alexander Grund, Alexander Lyashuk, Alexander Shaposhnikov, Alex Pivovarov, Aliia Khasanova, Alina Sbirlea, Allan Renucci, Amelia Thurdekoos, Amit Sabne, Andrei Ivanov, Andrew Dame, Andrey Portnoy, Anish Nair, Anlun Xu, Antonio Sanchez, anuj chincholikar, Anuj Chincholikar, Aravindh Balaji, aravindhbalaji1985, Arian Arfaian, Armin Felder, Artem Belevich, Ashish Rao, Ashitesh Singh, A. Unique TensorFlower, Bart Chrzaszcz, benediktjohannes, Benjamin Chetioui, Benjamin Kramer, Berkin Ilbeyi, Bhatu, Bhavani Subramanian, Bhupendra Dubey, Bill Varcho, Bixia Zheng, Blake Hechtman, Bodhi Silberling, BruceXinXin, Bryan Massoth, Buddh Prakash, Byungchul Kim, Ce Zheng, Changhui Lin, Chao, Charles Alaras, Chase Riley Roberts, Chenhao Jiang, Chris Ashton, Chris Jones, Chris Kennelly, Christian Sigg, Chuan He, Chunlei Niu, Chun-nien Chan, Chunyu Jin, Clive Verghese, Cong Liu, Corentin Kerisit, Daniel Chen, Daniel Kuts, Daniel Ng, Daniel Sosa, Daniel Suo, Danila Malyutin, David Dunleavy, David Majnemer, David Pizzuto, Deepika Rajani, deeptanshusekhri, dependabot[bot], Deqiang Chen, Derek Murray, Dillon Sharlet, Dimitar (Mitko) Asenov, Dimitris Vardoulakis, Dirk Hornung, DottsGit, Dragan Mladjenovic, Eetu Sjöblom, Elen Kalda, Emilio Cota, Emily Fertig, Eugene Zhulenev, Eusebio Durán Montaña, Evan Brown, Ezekiel Calubaquib, Faijul Amin, Felix Wang, Fengwu Yao, Fergus Henderson, Frederic Rechtenstein, Frederik Gossen, Gabriel Gerlero, Gagan Nagaraj, gaikwadrahul8, garry00107, gaurides, George Pawelczak, Georg Stefan Schmid, gns, Goran Flegar, Graham, Grant Jensen, Greg Olechwierowicz, Gregory Pataky, Grzegorz Gawryał, Gunhyun Park, guozhong.zhuang, Haibo Huang, Hana Joo, Hariprasad Ravishankar, Harsha H S, Harshit Monish, Henning Becker, Hittanshu, Hoeseong (Hayden) Kim, Hugo Mano, Hyeontaek Lim, Ibrahim Umit Akgun, ILCSFNO, Ilia Sergachev, Ilya Tikhonovskiy, Iman Hosseini, Ionel Gog, Isha Arkatkar, isharif168, Ivo Ristovski List, Jacques Pienaar, Jae H. Yoo, Jaeyoon Jung, Jake Harmon, James Hilliard, jameslovespancakes, James Spooner, Jane Liu, Jaroslav Sevcik, Jeff Parker, Jeffrey A. Dean, Jeremy Meredith, Jialei Chen, Jian Cai, Jian Li, Jie Luo, Jim Lin, Jing Pu, Jinliang Wei, Jiya Zhang, Joel Wee, Johannes Buchner, Johannes Reifferscheid, Johnny, Jorge Gorbe Moya, Joshua Lang, Joshua Wang, Joss Briody, jparkerh, Juanli Shen, Juhyun Lee, Jun Jiang, Junwhan Ahn, Kadir Barut, Kanglan Tang, Kanish Anand, Kanvi Khanna, Karlo Basioli, Ken Franko, Kevin Chen, Kevin Gleason, Kingston Mandisodza, Koki Ibukuro, Kostiantyn Liepieshov, Krishna Haridasan, Krishna Somani, Krzysztof Kosiński, Kuy Mainwaring, lambert, Larry Lansing, Lin Chai, Lord ε Rebel, Luke Baumann, Luke Hutton, madhavmadupu, Majid Dadashi, Mani Ananth, Manjunath Gaonkar, Marcello Maggioni, Marcin Radomski, Maria Lyubimtseva, Marissa Ikonomidis, Mark Daoust, Mason Ch

_[Truncated at 4000 characters — full notes: https://github.com/tensorflow/tensorflow/releases/tag/v2.21.0-rc0]_

### v2.20.0 — TensorFlow 2.20.0

- Date: 2025-08-13
- Version: v2.20.0
- Original notes: https://github.com/tensorflow/tensorflow/releases/tag/v2.20.0
- Permalink: https://whatsnew.fyi/product/tensorflow/releases/v2.20.0

- **changed** — The tensorflow-io-gcs-filesystem package is now optional and must be installed separately with pip install "tensorflow[gcs-filesystem]"
- **added** — Add autotune.min_parallelism to tf.data.Options to enable faster input pipeline warm up
- **deprecated** — tf.lite is deprecated in favor of the new LiteRT repository

#### Release 2.20.0

##### TensorFlow

###### Breaking Changes

* The `tensorflow-io-gcs-filesystem` package is now optional, due its uncertain, and limited support. To install it alongside `tensorflow`, run `pip install "tensorflow[gcs-filesystem]"`.

###### Major Features and Improvements

* `tf.data`
    * Adds `autotune.min_parallelism` to `tf.data.Options` to enable faster input pipeline warm up.
* `tf.lite`
    * tf.lite will be deprecated, in favor of the new repo https://github.com/google-ai-edge/LiteRT.
    * The duplicated source will also be removed from the TF repo.

##### Thanks to our Contributors

This release contains contributions from many people at Google, as well as:

1ndig0, 372046933, abhinav, afzpatel, Akhil Goel, Alain Carlucci, Aleksei, Alen Huang, Alex, Amrinfathima-Mcw, Aravindh Balaji, Armand Picard, Aseem Athale, Ashiq Imran, Assoap, Chao, Chase Riley Roberts, Chenhao Jiang, chunhsue, chuntl, Chunyu Jin, Corentin Kerisit, Crefeda Rodrigues, dependabot[bot], Dragan Mladjenovic, Elen Kalda, Felix Thomasmathibalan, gabeweisz, Gauri Deshpande, Georg Stefan Schmid, Guozhong Zhuang, Harsha H S, Harshith_N, Hugo Mano, Ian Tayler Lessa, Jack Wolfard, James Ward, Jane Liu, Jaroslav Sevcik, JD, Jerry-Ge, Jian Li, Jinzhe Zeng, jiunkaiy, Johannes Reifferscheid, johnnkp, junweifu, Kanvi Khanna, Kasper Nielsen, Linzb-Xyz, Luke Hutton, Mahmoud Abuzaina, Mathew Odden, Michael Platings, misterBart, Mitchell Ludwig, Mmakevic-Amd, mraunak, NamanAgarwal0905, Namrata-Ibm, Neuropilot-Captain, nhatle, Nicholas Wilson, Nikhil Shinde, Olli Lupton, Patrick J. Lopresti, Pavel Emeliyanenko, Pearu Peterson, pemeliya, Peng Sun, Philipp Hack, Pratham-Mcw, RahulSudarMCW, RakshithGB, Rakshithgb-Fujitsu, RuslanSemchenko, Ruturaj Vaidya, Sachin Muradi, sandeepgupta12, SaoirseARM, Sergey Kozub, Sevin Fide Varoglu, Shanbin Ke, Shaogang Wang, Shraiysh Vaishay, Siddhartha Menon, spiao, Swatheesh Muralidharan, Tai Ly, Terry Sun, Thibaut Goetghebuer-Planchon, Thomas Dickerson, Tilak, Tj Xu, Trevor Morris, tyb0807, vfdev, Wei Wang, wokron, wondertx, Xuefei Jiang, Yaowei Zhou, Zentrik, Ziyun Cheng, Zoranjovanovic-Ns

### v2.19.1 — TensorFlow 2.19.1

- Date: 2025-08-13
- Version: v2.19.1
- Original notes: https://github.com/tensorflow/tensorflow/releases/tag/v2.19.1
- Permalink: https://whatsnew.fyi/product/tensorflow/releases/v2.19.1

- **fixed** — Fix save_model.save for Serving embedding and add SparseCore Reshard

#### Release 2.19.1

###### Bug Fixes and Other Changes

* Fix save_model.save for Serving embedding and add SparseCore Reshard.

### v2.20.0-rc0 — TensorFlow 2.20.0-rc0

- Date: 2025-07-28
- Version: v2.20.0-rc0
- Original notes: https://github.com/tensorflow/tensorflow/releases/tag/v2.20.0-rc0
- Permalink: https://whatsnew.fyi/product/tensorflow/releases/v2.20.0-rc0
- Labels: Pre-release

- **removed** — The tensorflow-io-gcs-filesystem package is now optional and must be installed separately with pip install "tensorflow[gcs-filesystem]"
- **added** — Add autotune.min_parallelism to tf.data.Options to enable faster input pipeline warm up
- **deprecated** — tf.lite is deprecated in favor of the new repo https://github.com/google-ai-edge/LiteRT

#### Release 2.20.0

##### TensorFlow

###### Breaking Changes

* The `tensorflow-io-gcs-filesystem` package is now optional, due its uncertain, and limited support. To install it alongside `tensorflow`, run `pip install "tensorflow[gcs-filesystem]"`.

###### Major Features and Improvements

* `tf.data`
    * Adds `autotune.min_parallelism` to `tf.data.Options` to enable faster input pipeline warm up.
* `tf.lite`
    * tf.lite will be deprecated, in favor of the new repo https://github.com/google-ai-edge/LiteRT.
    * The duplicated source will also be removed from the TF repo.

##### Thanks to our Contributors

This release contains contributions from many people at Google, as well as:

1ndig0, 372046933, abhinav, afzpatel, Akhil Goel, Alain Carlucci, Aleksei, Alen Huang, Alex, Amrinfathima-Mcw, Aravindh Balaji, Armand Picard, Aseem Athale, Ashiq Imran, Assoap, Chao, Chase Riley Roberts, Chenhao Jiang, chunhsue, chuntl, Chunyu Jin, Corentin Kerisit, Crefeda Rodrigues, dependabot[bot], Dragan Mladjenovic, Elen Kalda, Felix Thomasmathibalan, gabeweisz, Gauri Deshpande, Georg Stefan Schmid, Guozhong Zhuang, Harsha H S, Harshith_N, Hugo Mano, Ian Tayler Lessa, Jack Wolfard, James Ward, Jane Liu, Jaroslav Sevcik, JD, Jerry-Ge, Jian Li, Jinzhe Zeng, jiunkaiy, Johannes Reifferscheid, johnnkp, junweifu, Kanvi Khanna, Kasper Nielsen, Linzb-Xyz, Luke Hutton, Mahmoud Abuzaina, Mathew Odden, Michael Platings, misterBart, Mitchell Ludwig, Mmakevic-Amd, mraunak, NamanAgarwal0905, Namrata-Ibm, Neuropilot-Captain, nhatle, Nicholas Wilson, Nikhil Shinde, Olli Lupton, Patrick J. Lopresti, Pavel Emeliyanenko, Pearu Peterson, pemeliya, Peng Sun, Philipp Hack, Pratham-Mcw, RahulSudarMCW, RakshithGB, Rakshithgb-Fujitsu, RuslanSemchenko, Ruturaj Vaidya, Sachin Muradi, sandeepgupta12, SaoirseARM, Sergey Kozub, Sevin Fide Varoglu, Shanbin Ke, Shaogang Wang, Shraiysh Vaishay, Siddhartha Menon, spiao, Swatheesh Muralidharan, Tai Ly, Terry Sun, Thibaut Goetghebuer-Planchon, Thomas Dickerson, Tilak, Tj Xu, Trevor Morris, tyb0807, vfdev, Wei Wang, wokron, wondertx, Xuefei Jiang, Yaowei Zhou, Zentrik, Ziyun Cheng, Zoranjovanovic-Ns

### v2.19.0 — TensorFlow 2.19.0

- Date: 2025-03-12
- Version: v2.19.0
- Original notes: https://github.com/tensorflow/tensorflow/releases/tag/v2.19.0
- Permalink: https://whatsnew.fyi/product/tensorflow/releases/v2.19.0

- **changed** — tflite::Interpreter public constants kTensorsReservedCapacity and kTensorsCapacityHeadroom are now const references instead of constexpr compile-time constants
- **deprecated** — tf.lite.Interpreter API will be deleted in TF 2.20, with migration to ai_edge_litert.interpreter
- **added** — tfl.Cast op now supports bfloat16 in runtime kernel
- **removed** — Stopped publishing libtensorflow packages (though it can still be unpacked from the PyPI package)

#### Release 2.19.0

##### TensorFlow

###### Breaking Changes

* `LiteRT`, a.k.a. `tf.lite`:
    * C++ API:
      * The public constants `tflite::Interpreter:kTensorsReservedCapacity` and `tflite::Interpreter:kTensorsCapacityHeadroom` are now const references, rather than `constexpr` compile-time constants. (This is to enable better API compatibility for TFLite in Play services while preserving the implementation flexibility to change the values of these constants in the future.)
    * Python API:
      * `tf.lite.Interpreter` gives deprecation warning redirecting to its new location at `ai_edge_litert.interpreter`, as the API `tf.lite.Interpreter` will be deleted in TF 2.20. See the [migration guide](https://ai.google.dev/edge/litert/migration) for details.

###### Known Caveats

###### Major Features and Improvements

*  `tf.lite`
    * `tfl.Cast` op is now supporting `bfloat16` in runtime kernel.

###### Bug Fixes and Other Changes

* We have stopped publishing `libtensorflow` packages but it can still be unpacked from the PyPI package.

##### Thanks to our Contributors

This release contains contributions from many people at Google, as well as:

Akhil Goel, akhilgoe, Alain Flaischer, Alex, Alexander Pivovarov, Alexander Shadchin, Alexis Praga, Amrinfathima-Mcw, Andrey Pikas, Andrey Portnoy, Ankur Singh, Ashiq Imran, Assoap, c8ef, charleshofer, Chase Riley Roberts, Chenhao Jiang, Chongyun Lee, Claudio Desouza, Corentin Godeau, Crefeda Rodrigues, Danny Burrow, dependabot[bot], Dimitris Vardoulakis, Dragan Mladjenovic, Elfie Guo, Emmanuel Ferdman, fiberflow, flyingcat, Gary Yi-Hung Chen, Georg Stefan Schmid, Gerwout Van Der Veen, Harsha H S, Harshit Monish, Hugo Mano, i.Pear, Ilia Sergachev, Jane Liu, Jaroslav Sevcik, Jc (Jonathan Chen), Jerry Ge, Jian Li, johndoknjas, Johnny, Jonathan Albrecht, Kaixi Hou, Kanvi Khanna, keerthanakadiri, Kevin Ji, Kiran Sai Ramineni, kwoncy2020, LakshmiKalaKadali, Lee, Jun Seok, Mahmoud Abuzaina, Matt Bahr, mayuyuace, Melissa Weber Mendonça, misterBart, Mkarpushin-Enhancelab, Mmakevic-Amd, mraunak, nallave, Nayana Thorat, Nayana-Ibm, nick.camarena, Nicolas Castet, Om Thakkar, oyzh, Parsa Homayouni, Patrick Toulme, Pavel Emeliyanenko, Pavithra Eswaramoorthy, Pearu Peterson, pemeliya, Philipp Hack, Ravi Kumar Soni, redwrasse, Ruturaj Vaidya, Sallenkey-Wei, Sandeep Gupta, Sergey Kozub, Sevin Fide Varoglu, Shanbin Ke, Shaogang Wang, Shixin Zhang, Shraiysh, Shu Wang, Silvio Traversaro, snadampal, Sunita Nadampalli, Tai Ly, Tatwai Chong, tchatow, tdanyluk, Terry Sun, Tilak, Tj Xu, Trevor Morris, Twice, vfdev, Vladimir Silyaev, Weisser, Pascal, wokron, Won Jeon, Xuefei Jiang, Zentrik, Zoranjovanovic-Ns

### v2.18.1 — TensorFlow 2.18.1

- Date: 2025-03-11
- Version: v2.18.1
- Original notes: https://github.com/tensorflow/tensorflow/releases/tag/v2.18.1
- Permalink: https://whatsnew.fyi/product/tensorflow/releases/v2.18.1

- **security** — Update curl to 8.11.0 to handle CVE-2024-2004, CVE-2024-2379, CVE-2024-2398, CVE-2024-2466, CVE-2024-6197, CVE-2024-7264, CVE-2024-8096 and CVE-2024-9681
- **changed** — Loosen ml_dtypes upperbound to < 1.0.0 to reduce conflicts when installed with other ML ecosystem components
- **deprecated** — tf.lite.Interpreter gives warning of future deletion and a redirection notice to its new location at ai_edge_litert.interpreter

#### Release 2.18.1

###### Security
* Updates curl to `8.11.0` to handle [CVE-2024-2004](https://github.com/advisories/GHSA-97xx-95pm-5qv6), [CVE-2024-2379](https://github.com/advisories/GHSA-wr4c-gwg7-p734), [CVE-2024-2398](https://github.com/advisories/GHSA-mq8w-c2j9-rqxc), [CVE-2024-2466](https://github.com/advisories/GHSA-9xr6-qf7m-2jv5), [CVE-2024-6197](https://github.com/advisories/GHSA-x3h8-3mf2-v794), [CVE-2024-7264](https://github.com/advisories/GHSA-97c4-2w4v-c7r8), [CVE-2024-8096](https://github.com/advisories/GHSA-gv3v-x3f3-7fxm) and [CVE-2024-9681](https://github.com/advisories/GHSA-g337-g667-mjvw).

###### Bug Fixes and Other Changes

* Loosen `ml_dtypes` upperbound to `< 1.0.0` to reduce conflicts when installed with other ML ecosystem components.  

###### Breaking Changes

* `tf.lite`
    * Interpreter:
      * `tf.lite.Interpreter` gives warning of future deletion and a redirection notice to its new location at `ai_edge_litert.interpreter`. See the [migration guide](https://ai.google.dev/edge/litert/migration) for details.
* Tensorflow-tpu for this patch is skipped due to some sparsecore related bugs. We suggest to upgrade to 2.19.0 instead.

### v2.19.0-rc0 — TensorFlow 2.19.0-rc0

- Date: 2025-02-24
- Version: v2.19.0-rc0
- Original notes: https://github.com/tensorflow/tensorflow/releases/tag/v2.19.0-rc0
- Permalink: https://whatsnew.fyi/product/tensorflow/releases/v2.19.0-rc0
- Labels: Pre-release

- **changed** — Public constants tflite::Interpreter::kTensorsReservedCapacity and tflite::Interpreter::kTensorsCapacityHeadroom are now const references rather than constexpr compile-time constants
- **deprecated** — tf.lite.Interpreter is deprecated and redirects to its new location at ai_edge_litert.interpreter, with deletion planned for TF 2.20
- **added** — tfl.Cast op now supports bfloat16 in runtime kernel
- **changed** — Stopped publishing libtensorflow packages but they can still be unpacked from the PyPI package

#### Release 2.19.0

##### TensorFlow

###### Breaking Changes

* `LiteRT`, a.k.a. `tf.lite`:
  * C++ API:
    * The public constants `tflite::Interpreter:kTensorsReservedCapacity` and `tflite::Interpreter:kTensorsCapacityHeadroom` are now const references, rather than `constexpr` compile-time constants. (This is to enable better API compatibility for TFLite in Play services while preserving the implementation flexibility to change the values of these constants in the future.)
    * Interpreter:
      * `tf.lite.Interpreter` gives deprecation warning redirecting to its new location at `ai_edge_litert.interpreter`, as the API `tf.lite.Interpreter` will be deleted in TF 2.20. See the [migration guide](https://ai.google.dev/edge/litert/migration) for details.

###### Known Caveats

###### Major Features and Improvements

*  `tf.lite`
    * `tfl.Cast` op is now supporting `bfloat16` in runtime kernel.

###### Bug Fixes and Other Changes

* We have stopped publishing `libtensorflow` packages but it can still be unpacked from the PyPI package.

##### Thanks to our Contributors

This release contains contributions from many people at Google, as well as:

Akhil Goel, akhilgoe, Alain Flaischer, Alex, Alexander Pivovarov, Alexander Shadchin, Alexis Praga, Amrinfathima-Mcw, Andrey Pikas, Andrey Portnoy, Ankur Singh, Ashiq Imran, Assoap, c8ef, charleshofer, Chase Riley Roberts, Chenhao Jiang, Chongyun Lee, Claudio Desouza, Corentin Godeau, Crefeda Rodrigues, Danny Burrow, dependabot[bot], Dimitris Vardoulakis, Dragan Mladjenovic, Elfie Guo, Emmanuel Ferdman, fiberflow, flyingcat, Gary Yi-Hung Chen, Georg Stefan Schmid, Gerwout Van Der Veen, Harsha H S, Harshit Monish, Hugo Mano, i.Pear, Ilia Sergachev, Jane Liu, Jaroslav Sevcik, Jc (Jonathan Chen), Jerry Ge, Jian Li, johndoknjas, Johnny, Jonathan Albrecht, Kaixi Hou, Kanvi Khanna, keerthanakadiri, Kevin Ji, Kiran Sai Ramineni, kwoncy2020, LakshmiKalaKadali, Lee, Jun Seok, Mahmoud Abuzaina, Matt Bahr, mayuyuace, Melissa Weber Mendonça, misterBart, Mkarpushin-Enhancelab, Mmakevic-Amd, mraunak, nallave, Nayana Thorat, Nayana-Ibm, nick.camarena, Nicolas Castet, Om Thakkar, oyzh, Parsa Homayouni, Patrick Toulme, Pavel Emeliyanenko, Pavithra Eswaramoorthy, Pearu Peterson, pemeliya, Philipp Hack, Ravi Kumar Soni, redwrasse, Ruturaj Vaidya, Sallenkey-Wei, Sandeep Gupta, Sergey Kozub, Sevin Fide Varoglu, Shanbin Ke, Shaogang Wang, Shixin Zhang, Shraiysh, Shu Wang, Silvio Traversaro, snadampal, Sunita Nadampalli, Tai Ly, Tatwai Chong, tchatow, tdanyluk, Terry Sun, Tilak, Tj Xu, Trevor Morris, Twice, vfdev, Vladimir Silyaev, Weisser, Pascal, wokron, Won Jeon, Xuefei Jiang, Zentrik, Zoranjovanovic-Ns

### v2.18.0 — TensorFlow 2.18.0

- Date: 2024-10-24
- Version: v2.18.0
- Original notes: https://github.com/tensorflow/tensorflow/releases/tag/v2.18.0
- Permalink: https://whatsnew.fyi/product/tensorflow/releases/v2.18.0

- **changed** — TfLiteOperatorCreate C API now accepts an optional fourth parameter
- **changed** — TensorRT support is disabled in CUDA builds for code health improvement
- **added** — Hermetic CUDA support added to use downloadable CUDA versions instead of locally installed CUDA
- **changed** — TensorFlow now supports and is compiled with NumPy 2.0 by default
- **added** — SignatureRunner is now supported for models with no signatures
- **added** — Add optional synchronous argument to tf.data.map to specify synchronous execution and save memory
- **added** — Add optional use_unbounded_threadpool argument to tf.data.map to use unbounded threadpool instead of core-based pool
- **added** — Add tf.data.experimental.get_model_proto to allow users to peek into the analytical model inside a dataset iterator
- **added** — Dequantize op supports TensorType_INT4 with per-channel dequantization
- **added** — Add support for stablehlo.composite in TensorFlow Lite
- **changed** — EmbeddingLookup op supports per-channel quantization and TensorType_INT4 values
- **changed** — FullyConnected op supports TensorType_INT16 activation and TensorType_Int4 weight per-channel quantization
- **added** — tf.tensor_scatter_update, tf.tensor_scatter_add and other reduce types support bad_indices_policy

#### Release 2.18.0

##### TensorFlow

###### Breaking Changes

* `tf.lite`
    * C API:
      * An optional, fourth parameter was added `TfLiteOperatorCreate` as a step forward towards a cleaner API for `TfLiteOperator`. Function `TfLiteOperatorCreate` was added recently, in TensorFlow Lite version 2.17.0, released on 7/11/2024, and we do not expect there will be much code using this function yet. Any code breakages can be easily resolved by passing nullptr as the new, 4th parameter.

* TensorRT support is disabled in CUDA builds for code health improvement.

* Hermetic CUDA support is added.

  Hermetic CUDA uses a specific downloadable version of CUDA instead of the user’s locally installed CUDA. Bazel will download CUDA, CUDNN and NCCL distributions, and then use CUDA libraries and tools as dependencies in various Bazel targets. This enables more reproducible builds for Google ML projects and supported CUDA versions. 

###### Known Caveats

###### Major Features and Improvements

* TensorFlow now supports and is compiled with NumPy 2.0 by default. Please see the [NumPy 2 release notes](https://numpy.org/doc/stable/release/2.0.0-notes.html) and the [NumPy 2 migration guide](https://numpy.org/devdocs/numpy_2_0_migration_guide.html#numpy-2-migration-guide).
   * Note that NumPy's type promotion rules have been changed(See [NEP 50](https://numpy.org/neps/nep-0050-scalar-promotion.html#nep50)for details). This may change the precision at which computations happen, leading either to type errors or to numerical changes to results.
   * Tensorflow will continue to support NumPy 1.26 until 2025, aligning with community standard deprecation timeline [here](https://scientific-python.org/specs/spec-0000/).
*   `tf.lite`:
    *   The LiteRT [repo](https://github.com/google-ai-edge/LiteRT) is live (see [announcement](https://developers.googleblog.com/en/tensorflow-lite-is-now-litert/)), which means that in the coming months there will be changes to the development experience for TFLite. The TF Lite Runtime source will be moved later this year, and sometime after that we will start accepting contributions through that repo.
* SignatureRunner is now supported for models with no signatures.

###### Bug Fixes and Other Changes

* `tf.data`
    * Add optional `synchronous` argument to `map`, to specify that the `map` should run synchronously, as opposed to be parallelizable when `options.experimental_optimization.map_parallelization=True`. This saves memory compared to setting `num_parallel_calls=1`.
    * Add optional `use_unbounded_threadpool` argument to `map`, to specify that the `map` should use an unbounded threadpool instead of the default pool that is based on the number of cores on the machine. This can improve throughput for map functions which perform IO or otherwise release the CPU.
    * Add [`tf.data.experimental.get_model_proto`](https://www.tensorflow.org/api_docs/python/tf/data/experimental/get_model_proto) to allow users to peek into the analytical model inside of a dataset iterator.

* `tf.lite`
    * `Dequantize` op supports `TensorType_INT4`.
        * This change includes per-channel dequantization.
    * Add support for `stablehlo.composite`.
    * `EmbeddingLookup` op supports per-channel quantization and `TensorType_INT4` values.
    * `FullyConnected` op supports `TensorType_INT16` activation and `TensorType_Int4` weight per-channel quantization.

* `tf.tensor_scatter_update`, `tf.tensor_scatter_add` and of other reduce types.
    * Support `bad_indices_policy`.

##### Thanks to our Contributors

This release contains contributions from many people at Google, as well as:

Akhil Goel, akhilgoe, Alexander Pivovarov, Amir Samani, Andrew Goodbody, Andrey Portnoy, Anthony Platanios, bernardoArcari, Brett Taylor, buptzyb, Chao, Christian Clauss, Cocoa, Daniil Kutz, Darya Parygina, dependabot[bot], Dimitris Vardoulakis, Dragan Mladjenovic, Elfie Guo, eukub, Faijul Amin, flyin

_[Truncated at 4000 characters — full notes: https://github.com/tensorflow/tensorflow/releases/tag/v2.18.0]_