Repository navigation
Tags: tensorflow/tensorflow
Tags
TensorFlow 2.22.0-rc0
* The TensorBoard dependency is no longer included by default. If you use `tf.summary.*` APIs or `tf.keras.callbacks.TensorBoard`, please install the `tensorboard` package separately (`pip install tensorboard`). Otherwise, TensorFlow will raise an ImportError.
* In `tensorflow/c/experimental/filesystem/filesystem_interface.h`, removed `TF_TransactionToken` and associated APIs.
* `tf.lite`
* Adds support for QUI4 (Quantized Unsigned 4-bit) in Dequantize operator.
* Adds support for FP16 and BF16 in Unpack operator.
* Adds support for FP16 in Transpose and DynamicUpdateSlice operator.
* Transpose now supports up to 8D tensors.
* Adds support for FLOAT8_E4M3FN and FLOAT8_E5M2 data types.
* `BatchFunction Operator`
* Adds the `num_warmup_batch_threads` op attribute to support a separate thread pool for processing warmup requests.
* Adds the `per_criticality_batch_timeout_micros` op attribute to support different batch timeouts for different criticality levels.
* `TensorFlow API`
* Exports `__new__` in public API golden files for subclasses of `tuple` (like `tf.io.FixedLenFeature`) to fix false positives during static type checking.
* `tf.data`
* Fixes a bug in `tf.data.Dataset.scan` where the shape of the state returned by `scan_func` was not strictly validated against the initial state.
* `tf.image.adjust_contrast`
* Registers the missing Python gradient for the `AdjustContrastv2` op, so `tf.image.adjust_contrast` can now be differentiated with `GradientTape`. Fixes [#126083](#126083).
* `tf.nn.softsign`
* Fixes second-order gradients of `tf.nn.softsign`. Differentiating twice previously failed with a lookup error because the `SoftsignGrad` backward op had no registered Python gradient.
* `tf.math.reciprocal`
* Constrains the XLA registration of `Reciprocal` and `Inv` to the types that have a device kernel, so `jit_compile=True` no longer silently accepts the integer inputs that eager execution and autoclustering reject. Fixes [#126414](#126414).
* `tf.experimental.numpy`
* `tf.experimental.numpy.isclose` and `tf.experimental.numpy.allclose` now apply `rtol` and `atol` to integer inputs, matching NumPy, instead of comparing with pure equality. The comparison stays in integer arithmetic; combining integer inputs with floating-point tolerances (including the defaults) now raises an error unless automatic type promotion is enabled, which keeps the cost of promotion opt-in.
* oneDNN (MKL) convolution and transpose kernels
* Raises `InvalidArgumentError` instead of aborting the process for a rank-mismatched `Conv3DBackpropFilterV2` input and for `ConjugateTranspose` on a scalar. Fixes [#118340](#118340) and [#118345](#118345).
* `tf.image.non_max_suppression`
* Fixes a segmentation fault under `jit_compile=True` when `boxes` and `scores` are empty. The XLA lowering now returns an empty selection, matching the non-XLA kernels. Fixes [#117245](#117245).
This release contains contributions from many people at Google, as well as:
0xASTRA, 3zzzz_0x, aarushisingh04, Abdul Wahab Ahmed Khan, Abhijeet Awasthi, abhijeet117, Abhinav Srivastava, Adi-Il, Adil Burak Şen, adilburaksen, Adithyan Ak, Aditya Gaurav, Ag Ramesh, Ahmed Sagar, Akhil Goel, Alex Y. Chan, Aman Sachan, andersendsa, animesh68, Ankit Vishwakarma, Anna Mayne, Anthony Platanios, Anupam Mediratta, Anupamkumar-1, Apoorv Darshan, AratiGanesh, Arun, Ashiq Imran, Ashish Rao, Ashitesh Singh, Ashutosh Kumar Singh, Ashutosh0x, Avneesh Jadhav, Ayaka Mikazuki, Beans, Benjamin Demaille, Bhavani Subramanian, Bradley Fargo, C. Antonio Sánchez, Chao, Chenna-Nagesh, Chenxi Hou, Chirag, Chirag Wattamwar, Christian Aurich, chuanzhisongshu, Chunyu Jin, clarke, Cocoa, CodeByMoriarty, codeXsidd, complexbulb, Corentin Kerisit, cuishuang, Cyrax321, Daksh Prajapati, Dakshprajapati-Afk, Daniil S, Danyal Ahmed, David Nichols, debaditya, Deep Patel, dependabot[bot], destro4evr, Dlcompilers-Infra-Bot, Doehyun Baek, Dragan Mladjenovic, Edrilan Berisha, Eetu Sjöblom, Elshad Toklayev, Emil Nichita, Emily, endorphin13, evilgensec, Eylon Eliyahu Krause, Faijul Amin, Florian Assmuth, foodlook, Furkan Egecan Nizam, FuZoe, gaurides, gladiator9797, Gulsum Gudukbay Akbulut, Guozhong Zhuang, Hanuman Singh, Harsha H S, Harsha Hs, Henry Bergin, Hsien-Ching Chung, htalaco, Hunterkritik-Byte, Huy Phung, Ian Tayler Lessa, Javid Khan, Jean-Baptiste Braun, Jian Li, Johannes Tenhumberg, Johnny Qiu, Justin Chiu, Kaivalya-Cyber, Kanvi Khanna, Kasper Nielsen, keerthanakadiri, Kejie Zhang, Kevin Barre, Keyur Ranipa, kgotlinux, kiwigitops, Kobi Hikri, kokol16, Kuldeeep18, Lakshit Verma, Lakshmi Arunachalam, Lao Fan, Lia, Linchen1-Robot, liruiyi, Luca Bruni, Maddipatla Chatan, madhanalagarsamy, Manish Reddy, Manjunath Gaonkar, Marco Minutoli, masaya.shiramizu, Matt Van Horn, Max Riffi-Aslett, Maxime France-Pillois, mcuiaws, Melissa Weber Mendonça, Mert Çakır, Michael Goldfarb, Milica Makevic, Min Htet Myet, Mina Ibrahim, Ming-Xu Huang, mohamedelabbas1996, Mohammadmseet-Hue, Mohammed Adib, mraunak, Mundur, Myself-Praveen, Nikhil Bhargava, Nikhita Hegde, Nishad Shabbir, Noxtimo, Nurul Gofran, nurulgofran, Oleg Eterevsky, Oliver Zhang, Olli Lupton, Om Thakkar, Omar Khaleel, Pavel Emeliyanenko, Pavithra Eswaramoorthy, pemeliya, Pham Binh, Praneeth Kodumagulla, Prasanna Dabi, prasanna8585, Priyanshu Doshi, Punya Modi, PwntGambit, qgao007, Qiu-Tiandev, Rachit-Mangawa, RadikHoroshev, Raghunandan Kumar, Rahul-S-007, Remin Franklin Eliyas, Ricardo Salveti, Rishu Ranjan, Roy Kim, rstar327, Ruturaj Kiran Vaidya, RuwanpuragePawan, Ryan Benney, Sahil Bhardwaj, Sai Ramineni, Sairam Veereddy, Saketh Reddy Pingili, Saksham Kapoor, Saksham Singh Rathore, Sam Ezeh, Samaresh Kumar Singh, SamareshSingh, Sankalp Thakur, Sepuri-Sai-Krishna, Sevin Fide Varoglu, Shanbin Ke, Shaogang Wang, Shin, Shinpache Shimura, shreyas, Som Tripathi, Srijan Upadhyay, Sudam Ranasinghe, Sumit Chauhan, Syed Mohammed Nayyar, Syed Osama Ali Shah, Techie03, Teerth Sharma, Terry Sun, Tirth, Tj Xu, tongke, Trighap52, Tristan, tsrw2048, Tushar Darote, Vaggelis, Vaisman, Vamshi Kiran Alugulla, Veer Jain, vfdev, Vibhor Gautam, VibhorGautam, Ville Vesilehto, VirajMishra1, Vishwak Thatikonda, Wei Wang, wilmerdooley, Xuefei Jiang, y198nt, Yash2277-Ctrl, Yasir Ashfaq, yasiribmcon, Yigtwxx, Yimei Sun, Zach, Zhihao "Zephyr" Yao, Zoranjovanovic-Ns
Backport Windows builds into 2.18 for 2.18.1 (#88848) * Update scripts/configs for Windows nightly/release builds. `set -u` (does not allow unbound variables) has been removed from all scripts. This is due to Docker on Windows treating variables in an env file, set to an empty value (`MY_VAR=`), as unbound variables. Consequently, these variables, even though they are "set", do not make it into the Docker container at all, and various checks for those variables fail outright. PiperOrigin-RevId: 713717958 (cherry picked from commit fa3b3c1) * Fix up the 2022 Win RBE config. PiperOrigin-RevId: 718920772 (cherry picked from commit 82ba59a) * Adjust the wheel.sh, libtensorflow.sh scripts to be Win-compatible for upload. Temporary workaround. PiperOrigin-RevId: 719027529 (cherry picked from commit 7513f77) * Add a workaround for gsutil not working properly on MSYS2. PiperOrigin-RevId: 719389902 (cherry picked from commit 9ab2fee) * Add the missing wheel name specification for Windows. PiperOrigin-RevId: 720176067 (cherry picked from commit 768059f) * Allow building identical wheels under additional names. Currently only needed for Windows, since the same wheel is uploaded to both tensorflow_cpu, and tensorflow PyPi repos, but different names/metadata are needed. PiperOrigin-RevId: 720313792 (cherry picked from commit a93dc78) * Actually copy the extra wheels into the expected directory afterwards. PiperOrigin-RevId: 720753829 (cherry picked from commit 6ba7e4e)
PreviousNext