- 01 7月, 2021 1 次提交
-
-
由 A. Unique TensorFlower 提交于
PiperOrigin-RevId: 382327442
-
- 25 6月, 2021 8 次提交
-
-
由 Chenkai Kuang 提交于
PiperOrigin-RevId: 381424814
-
由 TensorFlower Gardener 提交于
PiperOrigin-RevId: 381369501
-
由 A. Unique TensorFlower 提交于
PiperOrigin-RevId: 381368343
-
由 Yash Katariya 提交于
PiperOrigin-RevId: 381360615
-
由 Nick Felt 提交于
This simplifies the calling code, and ensures that we print all tags found the summary events if the assertion fails, which helps debug cases where summaries are still being logged but under a tag name that differs from the expected one. PiperOrigin-RevId: 381352128
-
由 Chenkai Kuang 提交于
PiperOrigin-RevId: 381292911
-
由 Katherine Wu 提交于
PiperOrigin-RevId: 381276605
-
由 TensorFlower Gardener 提交于
PiperOrigin-RevId: 381274594
-
- 24 6月, 2021 9 次提交
-
-
由 Evan Rosen 提交于
PiperOrigin-RevId: 381160816
-
由 Katherine Wu 提交于
This change also enables models with multiple input arguments to be saved, for example: ``` class Subclass(keras.Model): def call(self, a, b): ... ``` PiperOrigin-RevId: 381148805
-
由 Ruoxin Sang 提交于
PiperOrigin-RevId: 381088871
-
由 A. Unique TensorFlower 提交于
PiperOrigin-RevId: 381065953
-
由 A. Unique TensorFlower 提交于
PiperOrigin-RevId: 381061383
-
由 A. Unique TensorFlower 提交于
PiperOrigin-RevId: 381059294
-
由 A. Unique TensorFlower 提交于
PiperOrigin-RevId: 381058890
-
由 A. Unique TensorFlower 提交于
PiperOrigin-RevId: 381049091
-
由 Chenkai Kuang 提交于
ShardedVariable: add unit test to verify return value of assign_xxx can be passed as control dependencies. Add AUC metric test when variables are sharded. PiperOrigin-RevId: 381046405
-
- 23 6月, 2021 18 次提交
-
-
由 A. Unique TensorFlower 提交于
set `tf.expand_dims argument` `axis` to an integer instead of a list in the multi_head_attention model. PiperOrigin-RevId: 380957913
-
由 Evan Rosen 提交于
PiperOrigin-RevId: 380952115
-
由 Matt Watson 提交于
This was always broken for numpy vocabulary inputs, and recently broke for lists as well. PiperOrigin-RevId: 380946490
-
由 A. Unique TensorFlower 提交于
PiperOrigin-RevId: 380936605
-
由 Mark Daoust 提交于
1. was giving an `UnboundLocal` error. 2. If `fpath` and `untar_fpath` are set to the same value the `untar` doesn't run. So ensure that `fname` always has `.tar.gz`, and `untar_fpath` never does. The old behavior (before the previous change) was to always add `.tar.gz`, giving `.tar.gz.tar.gz` files. That seems wrong too. Fixes tensorflow_docs/site/en/tutorials/load_data/text.ipynb PiperOrigin-RevId: 380934634
-
由 Evan Rosen 提交于
PiperOrigin-RevId: 380933127
-
由 Matt Watson 提交于
PiperOrigin-RevId: 380918577
-
由 A. Unique TensorFlower 提交于
PiperOrigin-RevId: 380914431
-
由 A. Unique TensorFlower 提交于
PiperOrigin-RevId: 380914130
-
由 Matt Watson 提交于
It is only support for Discretization and Normalization and not even tested on those classes. Removing it gives us a cleaner API interface for release. PiperOrigin-RevId: 380906031
-
由 Matt Watson 提交于
We recently switched to creating our tables in an init_scope (which is good, we never want to creat them at graph execution time), but we also need to place the initialization ops inside the init_scope. PiperOrigin-RevId: 380899225
-
由 Scott Zhu 提交于
The date for forward compat already take effective. PiperOrigin-RevId: 380894653
-
由 Scott Zhu 提交于
PiperOrigin-RevId: 380894633
-
由 Yash Katariya 提交于
PiperOrigin-RevId: 380869424
-
由 Rick Chao 提交于
PiperOrigin-RevId: 380865941
-
由 Scott Zhu 提交于
PiperOrigin-RevId: 380851544
-
由 Luke Wood 提交于
- introduction to keras for engineers - introduction to keras for researchers PiperOrigin-RevId: 380842991
-
由 Scott Zhu 提交于
PiperOrigin-RevId: 380838692
-
- 22 6月, 2021 4 次提交
-
-
由 Francois Chollet 提交于
PiperOrigin-RevId: 380652823
-
由 Zirui Zhuang 提交于
Imported from GitHub PR https://github.com/tensorflow/tensorflow/pull/49292 Update metrics.py: Fix SparseCategoricalAccuracy.update_state() doc string Try fixing issue #49252 For SparseCategoricalAccuracy, `y_true` should be integer labels and `y_pred` should be probabilities. ``` m = tf.keras.metrics.SparseCategoricalAccuracy() m.update_state([[2], [1]], [[0.1, 0.6, 0.3], [0.05, 0.95, 0]]) # Correct usage, `y_true` as integer labels and `y_pred` as probabilities. print(m.result().numpy()) # >>> 0.5 m.update_state([[2], [1]], [[1], [1]]) # Wrong usage, both as integer labels. print(m.result().numpy()) # >>> 0.25 m.update_state([[0, 1, 0], [0, 1, 0]], [[0.1, 0.6, 0.3], [0.05, 0.95, 0]]) # Wrong usage, both as probabilities. print(m.result().numpy()) # >>> Error ``` Copybara import of the project: -- d6ee4a941b79a464a9b229f6a22be50baab62d4d by Zirui Zhuang <zr.zz.alp@gmail.com>: Update metrics.py Fix SparseCategoricalAccuracy.update_state() doc string -- 9410d39d0a9f850c4fb615a1d8b607f1b75d80c3 by Zirui Zhuang <zr.zz.alp@gmail.com>: Update tensorflow/python/keras/metrics.py correction on the description of y_true -- 57dbe0fd01d3d4548f3453013177e49073edb463 by Zirui Zhuang <zr.zz.alp@gmail.com>: Update tensorflow/python/keras/metrics.py explicitly return -- 8d83bffd4776bbfad12a4e2669e8e30f7a26d3a7 by Zirui Zhuang <zr.zz.alp@gmail.com>: Update tensorflow/python/keras/metrics.py correction on dimension annotation -- d064d6e41d2f3fad390e82eea6b4edc6b9c46558 by Zirui Zhuang <zr.zz.alp@gmail.com>: Update metrics.py reuse update_state docstring by using inheritance -- 2b7f4d924251a67a65953b1f83a0499986ee1475 by Zirui Zhuang <zr.zz.alp@gmail.com>: Revert "Update metrics.py" This reverts commit d064d6e41d2f3fad390e82eea6b4edc6b9c46558. -- 7e718e6538e72d125bc5edc857c15fc2053877a2 by Zirui Zhuang <zr.zz.alp@gmail.com>: Update metrics.py using private string UPDATE_STATE_DOCSTRING PiperOrigin-RevId: 380647901
-
由 Yash Katariya 提交于
PiperOrigin-RevId: 380646732
-
由 Matt Watson 提交于
This is a significant refactor of the internals of the layer, which will break SavedModel compatibility with previous versions. The usage of the layer will remain the same, so a compatible layer should be generatable from the same training script. This refactor has the following advantages: - Static tables can be distributed to end workers in a multi-worker setting allowing more efficient distributed training. - File based vocabularies will only be scanned once. - Static vocabularies passed on init will be consistently clonable with the layer config, rather than clonable only in the file based case. We now consistently enforce that a vocabulary must be set when calling the layer on anything besides a keras.Input. PiperOrigin-RevId: 380645230
-