| HEAD | 838c7da | Merge branch 'master' of github.com:keras-team/keras | 16 March 2024, 16:53:37 UTC |
| refs/heads/cherrypick | 8cb88fa | Merge branch 'r2.11' into cherrypick | 17 October 2022, 17:00:46 UTC |
| refs/heads/ctc | a42af06 | Fix code format | 19 December 2023, 21:37:09 UTC |
| refs/heads/functional_slicing | 2395c3b | Add softmax usage warning | 18 February 2024, 02:01:01 UTC |
| refs/heads/gradient_accumulation | 6ff2a72 | Revert test | 18 December 2023, 23:05:50 UTC |
| refs/heads/init_dtype | 1aa4496 | Fix typo. | 24 May 2019, 20:44:14 UTC |
| refs/heads/keras-1 | 141e05e | Update Theano installation instruction | 11 April 2016, 15:51:01 UTC |
| refs/heads/keras-2 | c45f48e | SWWAE Example: Conv kernel size in resid pathway to 1x1 and activation from BN+RELU to ELU (#5756) * Changed conv kernel size in resid pathway to 1x1, and changed activation from BN+RELU to ELU. * Added a more informative docstring decsribing elu argument and its two behaviors. | 15 March 2017, 05:54:25 UTC |
| refs/heads/lora | eba5a88 | Fix CI | 05 January 2024, 20:37:13 UTC |
| refs/heads/master | 838c7da | Merge branch 'master' of github.com:keras-team/keras | 16 March 2024, 16:53:37 UTC |
| refs/heads/mlx | cc2f446 | Adds initial support for an MLX backend (#18962) * Start the mlx backend * Add the numpy part of the backend and a native accessor * Start nn and fix core and trainer * Add mlx to the trainer test * Small fix in mlx/core * Fixes to pass the tests * Add floor in mlx backend * Fix the styles * Add several missing ops in MLX's nn * Fix formatting * Change mlx to return tuple shapes and add a few ops * Add definitions for more ops in the mlx backend * Add affine transformations * Add uint64 and int64 * Export rnn things from the backend init * Fix some errors * Fix diagonal indices and change exceptions to NYI * Enable the mlx backend in the tests * Update some things for mlx v0.2.0 * Formatter updates * Fix for python 3.9 * Add the Github workflow config * Add mlx in the import test * Fix import in mlx.math | 21 February 2024, 21:11:06 UTC |
| refs/heads/r2.10 | b80dd12 | Add `is_legacy_optimizer` to optimizer config to keep saving/loading consistent. (#16856) PiperOrigin-RevId: 463928027 Co-authored-by: Chen Qian <chenmoney@google.com> | 02 August 2022, 17:50:20 UTC |
| refs/heads/r2.11 | e6784e4 | Merge pull request #17225 from lgeiger:fix-mixed-precision-ema (#17226) PiperOrigin-RevId: 486999108 Co-authored-by: TensorFlower Gardener <gardener@tensorflow.org> | 08 November 2022, 22:37:12 UTC |
| refs/heads/r2.12 | f9336cc | Less restrictive fallback logic (#17591) Co-authored-by: Eugene Kuznetsov <eugene.kuznetsov@amd.com> | 23 February 2023, 18:12:17 UTC |
| refs/heads/r2.13 | b3ffea6 | Cherrypick Sequential serialization bug fix for r2.13 (#18258) | 27 June 2023, 17:22:22 UTC |
| refs/heads/r2.14 | 68f9af4 | Update requirements.txt (#18382) Update the tf version used for keras 2.14 rc. The pip package verification test was incorrectly using the tf-nightly and cause the build to fail. | 11 September 2023, 17:01:28 UTC |
| refs/heads/r2.15 | 601488f | Use TENSORFLOW_VERSION when available during pip_build script (#18739) | 06 November 2023, 23:55:08 UTC |
| refs/heads/r2.5 | 9c26610 | Sync OSS keras to https://github.com/tensorflow/tensorflow/commit/6af429790a6cf791d53a91d9fbf8a5a4b97da94b (where TF 2.5 RC is cut) Preparing for the Keras 2.5 RC cut. PiperOrigin-RevId: 365252577 | 26 March 2021, 17:38:56 UTC |
| refs/heads/r2.6 | 3a33d53 | Use the safer `safe_load` function instead of `unsafe_load` when possible There is no need to open ourselves up to arbitrary code execution, especially since this is not in a performance critical loop, so we can take the slowdown due to safety. PiperOrigin-RevId: 388501098 | 03 August 2021, 18:20:20 UTC |
| refs/heads/r2.6.1 | add7720 | Update training.py Fixes https://github.com/tensorflow/tensorflow/issues/52640 | 28 October 2021, 15:55:08 UTC |
| refs/heads/r2.8 | d8fcb9d | Compute LSTM and GRU via cuDNN for RaggedTensors. This cherry-picked commit fixes a bug preventing computing LSTM and GRU for RaggedTensors via cuDNN, resulting in a large speedup (easily 10 times). A TF2-only test comparing ragged and masked tensor LSTM and GRU on CPU and GPU is also provided. | 23 December 2021, 00:21:24 UTC |
| refs/heads/r2.9 | 07e1374 | Cherrypick ImageAugmetation related change to R2.9 (#16454) * Enable the keras dtensor API in OSS. PiperOrigin-RevId: 438858608 * Switching learning/brain dependency to OSS compatible test_util This is one test file failing, due to the monkey patching happens in the dtensor.init(), and I will need to dig more about the root cause (probably due to patching tf.Variable with DVariable, and cause logic difference for instance type checking.) PiperOrigin-RevId: 439676157 * Update the docstring for keras.dtensor components. 1. Add docstring for LayoutMap. 2. Hide certain methods for keras.dtensor.optimizers. PiperOrigin-RevId: 442651549 * add no-op label augmentation to RandomContrast PiperOrigin-RevId: 439938098 * Add augment_label to RandomBrightness PiperOrigin-RevId: 443122383 * Create augment_label method: ``` def augment_label(self, label, transformation=None): return label ``` on remaining image augmentation layers PiperOrigin-RevId: 443446864 * Revert "Create augment_label method:" This reverts commit 1694cb4ea0ca084c49a178afe45bbee4379a2cc4. Co-authored-by: Luke Wood <lukewood@google.com> | 22 April 2022, 17:34:55 UTC |
| refs/heads/stateless_model_loading | 492cbf1 | Reduce memory consumption for load_model (WiP) | 04 December 2023, 21:44:13 UTC |
| refs/heads/tf-keras | afbd5d3 | merge | 24 May 2017, 21:57:32 UTC |
| refs/tags/2.0.4 | c627fa5 | Prepare new PyPI release. | 29 April 2017, 23:18:54 UTC |
| refs/tags/2.0.5 | d4b618b | Prepare new PyPI release. | 12 June 2017, 18:46:21 UTC |
| refs/tags/2.0.6 | f120a56 | Prepare 2.0.6 release. | 07 July 2017, 20:48:29 UTC |
| refs/tags/2.0.7 | 14a12e9 | Prepare 2.0.7 release. | 21 August 2017, 22:13:50 UTC |
| refs/tags/2.0.8 | c7dd36a | Prepare 2.0.8 release. | 25 August 2017, 19:01:46 UTC |
| refs/tags/2.0.9 | c25e06f | Prepare 2.0.9 release | 01 November 2017, 20:03:25 UTC |
| refs/tags/2.1.0 | e3e97c4 | Prepare new PyPI release. | 13 November 2017, 19:52:31 UTC |
| refs/tags/2.1.1 | 018e55b | Prepare PyPI release. | 14 November 2017, 21:40:18 UTC |
| refs/tags/2.1.2 | 26d5e10 | Prepare 2.1.2 release | 01 December 2017, 18:12:33 UTC |
| refs/tags/2.1.3 | 13402f1 | Prepare new PyPI release | 16 January 2018, 04:55:08 UTC |
| refs/tags/2.1.4 | 5d54eeb | Prepare 2.1.4 release. | 13 February 2018, 22:52:55 UTC |
| refs/tags/2.1.5 | 2fae461 | Prepare 2.1.5 release | 06 March 2018, 22:05:39 UTC |
| refs/tags/2.1.6 | b09ec1c | fixing typos (#10016) | 23 April 2018, 19:43:48 UTC |
| refs/tags/2.2.0 | 632560d | Style fix | 06 June 2018, 18:53:18 UTC |
| refs/tags/2.2.1 | eb2c1fd | Prepare 2.2.1 release. | 27 July 2018, 22:06:51 UTC |
| refs/tags/2.2.2 | 54386ef | Prepare new release. | 28 July 2018, 19:32:13 UTC |
| refs/tags/2.2.3 | ecbf73f | [RELNOTES] [P] Write to TensorBoard every x samples. (#11152) * Working on improving tensor flow callbacks * Adding batch level TensorBoard logging (implementing the `on_batch_end` method to the TensorBoard class * Interim commit -- added notes. * Corrected stylistic issues -- brought to compliance w/ PEP8 * Added the missing argument in the test suite. * Added the possibility to choose how frequently tensorboard should log the metrics and losses. * Fixed the issue of the validation data not being displayed. * Fixed the issue about the callback not remembering when was the last time it wrote to the logs. * Removed the error check. * Used update_freq instead of write_step. * Forgot to change the constructor call. | 01 October 2018, 19:00:01 UTC |
| refs/tags/2.2.4 | 1931e21 | Prepare 2.2.4 release. | 03 October 2018, 20:58:20 UTC |
| refs/tags/2.2.5 | fb7f49e | Fix sklearn regressor test? | 22 August 2019, 00:58:08 UTC |
| refs/tags/2.3.0 | a0335a3 | Update README | 17 September 2019, 17:01:48 UTC |
| refs/tags/2.3.1 | f295e8e | Prepare 2.3.1 release | 07 October 2019, 20:03:26 UTC |
| refs/tags/2.4.0 | b5cb82c | Redirect Keras to tf.keras (#14121) * Replace all code with tf.keras redirects | 17 June 2020, 21:32:46 UTC |
| refs/tags/v2.10.0 | b80dd12 | Add `is_legacy_optimizer` to optimizer config to keep saving/loading consistent. (#16856) PiperOrigin-RevId: 463928027 Co-authored-by: Chen Qian <chenmoney@google.com> | 02 August 2022, 17:50:20 UTC |
| refs/tags/v2.10.0-rc1 | b80dd12 | Add `is_legacy_optimizer` to optimizer config to keep saving/loading consistent. (#16856) PiperOrigin-RevId: 463928027 Co-authored-by: Chen Qian <chenmoney@google.com> | 02 August 2022, 17:50:20 UTC |
| refs/tags/v2.11.0 | e6784e4 | Merge pull request #17225 from lgeiger:fix-mixed-precision-ema (#17226) PiperOrigin-RevId: 486999108 Co-authored-by: TensorFlower Gardener <gardener@tensorflow.org> | 08 November 2022, 22:37:12 UTC |
| refs/tags/v2.11.0-rc1 | b12b9a1 | Cherrypick 2 more changes for the optimizer docstring fix. (#17165) * keras: fix typo in keras optimizers migration recommendation tf.keras.optimizer.legacy -> tf.keras.optimizers.legacy PiperOrigin-RevId: 481720094 * Keras: Fix docstring for tf.keras.optimizer. -> tf.keras.optimizers. PiperOrigin-RevId: 481721074 Co-authored-by: Michael Chinen <mchinen@google.com> Co-authored-by: A. Unique TensorFlower <gardener@tensorflow.org> | 17 October 2022, 21:18:54 UTC |
| refs/tags/v2.11.0-rc2 | 7411db0 | Cherrypick for cl/482011499: Throw error on deprecated fields. (#17179) * Throw error on deprecated fields. PiperOrigin-RevId: 482011499 * copyedits Co-authored-by: Chen Qian <chenmoney@google.com> Co-authored-by: François Chollet <francois.chollet@gmail.com> | 24 October 2022, 18:53:39 UTC |
| refs/tags/v2.11.0-rc3 | e6784e4 | Merge pull request #17225 from lgeiger:fix-mixed-precision-ema (#17226) PiperOrigin-RevId: 486999108 Co-authored-by: TensorFlower Gardener <gardener@tensorflow.org> | 08 November 2022, 22:37:12 UTC |
| refs/tags/v2.12.0 | f9336cc | Less restrictive fallback logic (#17591) Co-authored-by: Eugene Kuznetsov <eugene.kuznetsov@amd.com> | 23 February 2023, 18:12:17 UTC |
| refs/tags/v2.12.0-rc1 | f9336cc | Less restrictive fallback logic (#17591) Co-authored-by: Eugene Kuznetsov <eugene.kuznetsov@amd.com> | 23 February 2023, 18:12:17 UTC |
| refs/tags/v2.13.1 | b3ffea6 | Cherrypick Sequential serialization bug fix for r2.13 (#18258) | 27 June 2023, 17:22:22 UTC |
| refs/tags/v2.13.1-rc0 | 87db506 | Cherrypick the release script fix for RC. (#18082) | 02 May 2023, 18:00:30 UTC |
| refs/tags/v2.13.1-rc1 | b3ffea6 | Cherrypick Sequential serialization bug fix for r2.13 (#18258) | 27 June 2023, 17:22:22 UTC |
| refs/tags/v2.14.0 | 68f9af4 | Update requirements.txt (#18382) Update the tf version used for keras 2.14 rc. The pip package verification test was incorrectly using the tf-nightly and cause the build to fail. | 11 September 2023, 17:01:28 UTC |
| refs/tags/v2.14.0-rc0 | bb9d7ff | Removes Python 3.8 support (#18332) | 02 August 2023, 22:27:22 UTC |
| refs/tags/v2.15.0 | 601488f | Use TENSORFLOW_VERSION when available during pip_build script (#18739) | 06 November 2023, 23:55:08 UTC |
| refs/tags/v2.6.0 | 3a33d53 | Use the safer `safe_load` function instead of `unsafe_load` when possible There is no need to open ourselves up to arbitrary code execution, especially since this is not in a performance critical loop, so we can take the slowdown due to safety. PiperOrigin-RevId: 388501098 | 03 August 2021, 18:20:20 UTC |
| refs/tags/v2.6.0-rc0 | 71d6d6e | Add ParameterServerStrategy combination to ketrics_metrics_test. PiperOrigin-RevId: 381424814 | 25 June 2021, 08:47:51 UTC |
| refs/tags/v2.6.0-rc1 | d1e2589 | Internal change PiperOrigin-RevId: 382327442 | 30 June 2021, 16:58:23 UTC |
| refs/tags/v2.6.0-rc2 | 433eaa0 | Add more explicit error message for instance type checking of optimizer. PiperOrigin-RevId: 382409145 | 30 June 2021, 23:30:30 UTC |
| refs/tags/v2.6.0-rc3 | 3a33d53 | Use the safer `safe_load` function instead of `unsafe_load` when possible There is no need to open ourselves up to arbitrary code execution, especially since this is not in a performance critical loop, so we can take the slowdown due to safety. PiperOrigin-RevId: 388501098 | 03 August 2021, 18:20:20 UTC |
| refs/tags/v2.7,0 | 2c48a3b | Disable the failing tests due to numpy 1.20 change (#15552) The test has been disabled on master branch as well, we need to cherrypick this to unblock the release build. original cl in cl/400819173. | 27 October 2021, 00:56:39 UTC |
| refs/tags/v2.7.0 | 2c48a3b | Disable the failing tests due to numpy 1.20 change (#15552) The test has been disabled on master branch as well, we need to cherrypick this to unblock the release build. original cl in cl/400819173. | 27 October 2021, 00:56:39 UTC |
| refs/tags/v2.7.0-rc0 | 89fd0f5 | Remove temporary monitoring now that underlying perf issue is resolved PiperOrigin-RevId: 398533606 | 23 September 2021, 18:17:22 UTC |
| refs/tags/v2.7.0-rc1 | 87c5f61 | Update Keras Functional model docstring for the new feature of submodel slicing. PiperOrigin-RevId: 399303217 | 27 September 2021, 23:04:03 UTC |
| refs/tags/v2.7.0-rc2 | 2c48a3b | Disable the failing tests due to numpy 1.20 change (#15552) The test has been disabled on master branch as well, we need to cherrypick this to unblock the release build. original cl in cl/400819173. | 27 October 2021, 00:56:39 UTC |
| refs/tags/v2.8.0 | d8fcb9d | Compute LSTM and GRU via cuDNN for RaggedTensors. This cherry-picked commit fixes a bug preventing computing LSTM and GRU for RaggedTensors via cuDNN, resulting in a large speedup (easily 10 times). A TF2-only test comparing ragged and masked tensor LSTM and GRU on CPU and GPU is also provided. | 23 December 2021, 00:21:24 UTC |
| refs/tags/v2.8.0-rc0 | e812c17 | Fix RandomCrop layer called on integer input We had a tf.cond which called tf.image.resize on one path (returning float32) and cropping the input dtype on the other, which lead to a error. Instead we cast the inputs to the layer's compute dtype and do our computation with that type. Fixes https://github.com/keras-team/keras/issues/15681 PiperOrigin-RevId: 417686461 | 21 December 2021, 22:01:39 UTC |
| refs/tags/v2.8.0-rc1 | d8fcb9d | Compute LSTM and GRU via cuDNN for RaggedTensors. This cherry-picked commit fixes a bug preventing computing LSTM and GRU for RaggedTensors via cuDNN, resulting in a large speedup (easily 10 times). A TF2-only test comparing ragged and masked tensor LSTM and GRU on CPU and GPU is also provided. | 23 December 2021, 00:21:24 UTC |
| refs/tags/v2.9.0 | 07e1374 | Cherrypick ImageAugmetation related change to R2.9 (#16454) * Enable the keras dtensor API in OSS. PiperOrigin-RevId: 438858608 * Switching learning/brain dependency to OSS compatible test_util This is one test file failing, due to the monkey patching happens in the dtensor.init(), and I will need to dig more about the root cause (probably due to patching tf.Variable with DVariable, and cause logic difference for instance type checking.) PiperOrigin-RevId: 439676157 * Update the docstring for keras.dtensor components. 1. Add docstring for LayoutMap. 2. Hide certain methods for keras.dtensor.optimizers. PiperOrigin-RevId: 442651549 * add no-op label augmentation to RandomContrast PiperOrigin-RevId: 439938098 * Add augment_label to RandomBrightness PiperOrigin-RevId: 443122383 * Create augment_label method: ``` def augment_label(self, label, transformation=None): return label ``` on remaining image augmentation layers PiperOrigin-RevId: 443446864 * Revert "Create augment_label method:" This reverts commit 1694cb4ea0ca084c49a178afe45bbee4379a2cc4. Co-authored-by: Luke Wood <lukewood@google.com> | 22 April 2022, 17:34:55 UTC |
| refs/tags/v2.9.0-rc0 | 55476a8 | Change keras to use dtensor public API. PiperOrigin-RevId: 438605222 | 31 March 2022, 17:38:57 UTC |
| refs/tags/v2.9.0-rc1 | 27e3966 | Cherrypick Keras DTensor related updates into keras 2.9 (#16379) * Enable the keras dtensor API in OSS. PiperOrigin-RevId: 438858608 * Switching learning/brain dependency to OSS compatible test_util This is one test file failing, due to the monkey patching happens in the dtensor.init(), and I will need to dig more about the root cause (probably due to patching tf.Variable with DVariable, and cause logic difference for instance type checking.) PiperOrigin-RevId: 439676157 | 07 April 2022, 23:23:08 UTC |
| refs/tags/v2.9.0-rc2 | 07e1374 | Cherrypick ImageAugmetation related change to R2.9 (#16454) * Enable the keras dtensor API in OSS. PiperOrigin-RevId: 438858608 * Switching learning/brain dependency to OSS compatible test_util This is one test file failing, due to the monkey patching happens in the dtensor.init(), and I will need to dig more about the root cause (probably due to patching tf.Variable with DVariable, and cause logic difference for instance type checking.) PiperOrigin-RevId: 439676157 * Update the docstring for keras.dtensor components. 1. Add docstring for LayoutMap. 2. Hide certain methods for keras.dtensor.optimizers. PiperOrigin-RevId: 442651549 * add no-op label augmentation to RandomContrast PiperOrigin-RevId: 439938098 * Add augment_label to RandomBrightness PiperOrigin-RevId: 443122383 * Create augment_label method: ``` def augment_label(self, label, transformation=None): return label ``` on remaining image augmentation layers PiperOrigin-RevId: 443446864 * Revert "Create augment_label method:" This reverts commit 1694cb4ea0ca084c49a178afe45bbee4379a2cc4. Co-authored-by: Luke Wood <lukewood@google.com> | 22 April 2022, 17:34:55 UTC |
| refs/tags/v3.0.0 | 9c675a9 | Bug fix for TF backend using random zoom with flatten (#18835) * bug fix * skip the numpy backend | 27 November 2023, 23:53:15 UTC |
| refs/tags/v3.0.1 | 037ec9f | Fix code format. | 06 December 2023, 19:41:01 UTC |
| refs/tags/v3.0.2 | fe2f54a | Correct the Error description in cropping layer APIs (#18974) * Correct the Error description in cropping APIs * Fix failing cropping test cases | 21 December 2023, 06:10:45 UTC |
| refs/tags/v3.0.3 | 096b848 | Allow bfloat16 default dtype (#19074) Useful for llms! The tradeoff in precision can often be worth it in memory constrained environments, and unlike float16, does not have the same overflow/underflow issues during training. | 20 January 2024, 01:36:07 UTC |
| refs/tags/v3.0.4 | 19187d8 | Merge branch 'master' of github.com:keras-team/keras | 20 January 2024, 19:40:40 UTC |
| refs/tags/v3.0.5 | e6e6240 | Update version number. | 14 February 2024, 21:50:35 UTC |
| refs/tags/v3.1.0 | 0d0be6a | Add set_backend() utility. | 11 March 2024, 13:54:49 UTC |
