- 03 11月, 2020 1 次提交
-
-
由 Jay Shi 提交于
PiperOrigin-RevId: 340252424 Change-Id: Ic5a5d347c38fd94b68e10373c0b66ea3043f2e9b
-
- 27 10月, 2020 2 次提交
- 20 10月, 2020 1 次提交
-
-
由 Jay Shi 提交于
PiperOrigin-RevId: 337891662 Change-Id: Iaeaf7c94fc6ab4ee880b5ff5dc2a8e04c7ca0af9
-
- 13 10月, 2020 1 次提交
-
-
由 Jay Shi 提交于
PiperOrigin-RevId: 336717806 Change-Id: I724c31d9d2cb4cff1c867c168e52aa026c05c01f
-
- 07 10月, 2020 1 次提交
-
-
由 Jay Shi 提交于
PiperOrigin-RevId: 335747562 Change-Id: I9ce62981d540131f9a6297f61eaea1b6e0def492
-
- 06 10月, 2020 1 次提交
-
-
由 Jay Shi 提交于
[tf.data] Stop to record graduated experiments for in the future some of them may not appear in the `graduated_experiments` list. PiperOrigin-RevId: 335495276 Change-Id: Ibf11b782371375ebc6acd7e6a906d68fa0128bae
-
- 25 9月, 2020 1 次提交
-
-
由 Jiri Simsa 提交于
PiperOrigin-RevId: 333597626 Change-Id: I63be2d4b736d7d1423c93999a0ac2184257eadb4
-
- 23 9月, 2020 1 次提交
-
-
由 Jay Shi 提交于
PiperOrigin-RevId: 333121408 Change-Id: I6e75f7ba971c591346edcc7bcb92e0c16e84e8df
-
- 15 9月, 2020 1 次提交
-
-
由 Jay Shi 提交于
PiperOrigin-RevId: 331566266 Change-Id: Id5c754caa638b5ba7b841994607f25386b43ef21
-
- 09 9月, 2020 1 次提交
-
-
由 Jay Shi 提交于
PiperOrigin-RevId: 330537696 Change-Id: I757b3d7c1e321502fc3f145fb008f8d3fb55cb40
-
- 02 9月, 2020 1 次提交
-
-
由 Jay Shi 提交于
PiperOrigin-RevId: 329538533 Change-Id: I67d38f4c7cc59cc1914d42c07d059fed71e9e334
-
- 25 8月, 2020 1 次提交
-
-
由 Jay Shi 提交于
PiperOrigin-RevId: 328182943 Change-Id: I5e5784a9185742f2f1d114d69f23c3d1da412b3f
-
- 19 8月, 2020 1 次提交
-
-
由 Jay Shi 提交于
[tf.data] Only apply the optimization `disable_intra_op_parallelism` on the main dataset. Also update the unit test. PiperOrigin-RevId: 327304127 Change-Id: I19cd40a7633a52914ad438cda7b2608c1d835413
-
- 15 8月, 2020 1 次提交
-
-
由 Jiri Simsa 提交于
PiperOrigin-RevId: 326673260 Change-Id: Ie7be47834d827a6db61ed1e46f247bc65e4b2fcc
-
- 14 8月, 2020 1 次提交
-
-
由 Jay Shi 提交于
PiperOrigin-RevId: 326507486 Change-Id: Ic5b8b9cc67d327f3f0b55e1dfb28cd0edc882e7d
-
- 13 8月, 2020 1 次提交
-
-
由 Jay Shi 提交于
[tf.data] Add `OptimizeDatasetOp` to most input pipelines. Enable graph rewrites only when there are optimizations to be applied to the pipelines. PiperOrigin-RevId: 326312484 Change-Id: Ifdf1274db7fd8f07bb6c1815152f6f5938797de2
-
- 11 8月, 2020 1 次提交
-
-
由 Jay Shi 提交于
[tf.data] Add unit test to test the environment variable settings in `ObtainOptimizations` function. PiperOrigin-RevId: 325903283 Change-Id: I45742135cf4afe7c45c29afd9585b0445c07dd0a
-
- 08 8月, 2020 1 次提交
-
-
由 A. Unique TensorFlower 提交于
PiperOrigin-RevId: 325458185 Change-Id: I38905998dab7bfb5802ad94689282db4c08e271e
-
- 07 8月, 2020 1 次提交
-
-
由 Jay Shi 提交于
PiperOrigin-RevId: 325345128 Change-Id: I8de30c47f681a6f41e25e5ade4460f20a9d7bb5d
-
- 28 7月, 2020 3 次提交
-
-
由 A. Unique TensorFlower 提交于
PiperOrigin-RevId: 323507899 Change-Id: I5ede4ed67d92f132a4e8b7fbe175084daa4181ec
-
由 A. Unique TensorFlower 提交于
PiperOrigin-RevId: 323493103 Change-Id: I763bacdc55636287304f8f4f0f3a3acbe89e4198
-
由 Jay Shi 提交于
PiperOrigin-RevId: 323475203 Change-Id: I4fc60053deda12b1536c17a2dfe941cf9113aed6
-
- 06 6月, 2020 1 次提交
-
-
由 Rachel Lim 提交于
Handle DeadlineExceeded in OptimizeDatasetOp::MakeDataset correctly. Before this change, if we encounter a DeadlineExceeded error, the output dataset will not be set (i.e. be a nullptr), resulting in "Read uninitialized Dataset variant" when we try to use it. PiperOrigin-RevId: 315005357 Change-Id: I91f1f3b4e99ddc733ec11b5040c6d01b094422ba
-
- 15 4月, 2020 1 次提交
-
-
由 Jiri Simsa 提交于
This CL: 1) Introduces a mechanism for identifying function graphs associated with tf.data user defined functions and uses it to exclude such graphs from Grappler optimizations applied by TensorFlow optimizer to the entire function library runtime (in graph mode). 2) Extends the set of Grappler optimizations applied by tf.data optimizer to function graphs associated with tf.data user-defined functions to match the default TensorFlow optimizer. 3) Introduces a new tf.data rewrite, which sets the device of Conv2D without explicit device assignment to CPU. The layout optimization assumes that all ops without explicit device assignment will be placed on GPU (if possible), which is not true for ops without tf.data user-defined function and result in incorrect layout optimization. PiperOrigin-RevId: 306538159 Change-Id: I963831533b462dc1fb96406caa811c082cdb7125
-
- 22 2月, 2020 1 次提交
-
-
由 Jiri Simsa 提交于
[tf.data] DeadlineExceededError from input pipeline graph rewrites should not prevent further computation. PiperOrigin-RevId: 296513646 Change-Id: I9492c986567502f52bce63c04f9f7bb6d9da76ac
-
- 10 10月, 2019 1 次提交
-
-
由 Jiri Simsa 提交于
PiperOrigin-RevId: 273811032
-
- 04 10月, 2019 2 次提交
-
-
由 Jiri Simsa 提交于
PiperOrigin-RevId: 272770726
-
由 Jiri Simsa 提交于
PiperOrigin-RevId: 272740411
-
- 23 8月, 2019 1 次提交
-
-
由 Derek Murray 提交于
This simplifies the task of building tf.data on platforms that cannot build all of the Grappler dependencies. PiperOrigin-RevId: 265056563
-
- 17 8月, 2019 1 次提交
-
-
由 Dero Gharibian 提交于
Removed superfluous USE_TSTRING guards in core/framework/dataset.h. This is a part of a larger migration effort for tensorflow::tstring. See: https://github.com/tensorflow/community/pull/91 PiperOrigin-RevId: 263798289
-
- 16 7月, 2019 1 次提交
-
-
由 Fei Hu 提交于
-
- 04 7月, 2019 1 次提交
-
-
由 A. Unique TensorFlower 提交于
PiperOrigin-RevId: 256549401
-
- 04 6月, 2019 1 次提交
-
-
由 Rachel Lim 提交于
[tf.data] Have experimental_slack fail on nonconforming input pipelines. Also updates optimize_dataset_op to fail if any of the optimizers fail (previously, it would log an error but continue). PiperOrigin-RevId: 251259335
-
- 03 5月, 2019 2 次提交
-
-
由 Jiri Simsa 提交于
PiperOrigin-RevId: 246502904
-
由 Derek Murray 提交于
PiperOrigin-RevId: 246349191
-
- 02 5月, 2019 1 次提交
-
-
由 Jiri Simsa 提交于
[tf.data] Convert `GraphRewriteDataset` to a utility method as it no longer needs to maintain state after the introduction of `FunctionMetadata` and runtime fixes. PiperOrigin-RevId: 246237582
-
- 20 3月, 2019 1 次提交
-
-
由 Rachel Lim 提交于
[tf.data] Add option to control whether vectorization is aggressive (i.e. always vectorizes) or safe (i.e. uses ChooseFastestBranchDataset) PiperOrigin-RevId: 239203789
-
- 02 2月, 2019 1 次提交
-
-
由 Rohan Jain 提交于
Adding a RebatchDatasetOp that uses Grappler to modify the dataset graph to mutate the batch size of the Dataset. Right now this basically does all the wiring through Grappler etc. and the logic is very minimal and basic. Subsequent CL's will make the logic expand to different scenarios. PiperOrigin-RevId: 231983575
-
- 15 1月, 2019 1 次提交
-
-
由 Jiri Simsa 提交于
This CL: - adds counters for tf.data elements, autotuning and optimizations - sets the number of iterations of the `tf_data_meta_optimizer` to one -- the iteration of tf.data optimizations is handled by the tf.data meta optimizer itself - adds the `alwayslink` attribute to all tf.data optimization BUILD targets to make sure they are always registered (without this, they would not be registered for the Tensorflow server binary I was using for local testing) and further cleans up visibility and dependencies of //third_party/tensorflow/core/grappler/optimizers/data/BUILD - introduces TFDataOptimizerBase as a base class for tf.data optimizations - moves TensorFlow metrics into tensorflow::metrics namespace PiperOrigin-RevId: 229302097
-