https://github.com/deepmind/sonnet

sort by:
Revision Author Date Message Commit Date
ee1731e Sonnet version update produced on Friday, 06. September 2019 PiperOrigin-RevId: 267584068 06 September 2019, 13:21:10 UTC
3d57a44 [Automated] Update package documentation. PiperOrigin-RevId: 267555272 06 September 2019, 13:21:01 UTC
2b456dc [Automated] Update package documentation. PiperOrigin-RevId: 266882919 06 September 2019, 13:20:51 UTC
10e4332 [Automated] Update package documentation. PiperOrigin-RevId: 266333702 06 September 2019, 13:20:41 UTC
eea3e2a [Automated] Update package documentation. PiperOrigin-RevId: 265648268 06 September 2019, 13:20:32 UTC
d5e2dc4 [Automated] Update package documentation. PiperOrigin-RevId: 265029581 06 September 2019, 13:20:23 UTC
defdf30 Refresh dead link to TF API docs. PiperOrigin-RevId: 264840833 06 September 2019, 13:20:13 UTC
de7d109 Switch tf.GraphKeys -> tf.compat.v1.GraphKeys PiperOrigin-RevId: 264573324 06 September 2019, 13:20:05 UTC
4311b4c [Automated] Update package documentation. PiperOrigin-RevId: 264345902 06 September 2019, 13:19:56 UTC
19a6e1c [Automated] Update package documentation. PiperOrigin-RevId: 263732206 06 September 2019, 13:19:46 UTC
7a2d068 Bug fixes. PiperOrigin-RevId: 263619074 06 September 2019, 13:19:38 UTC
cd278c5 [Automated] Update package documentation. PiperOrigin-RevId: 263561993 06 September 2019, 13:19:28 UTC
7a33bd2 Internal fixes. PiperOrigin-RevId: 261315452 06 September 2019, 13:19:19 UTC
a3fabf6 Add code for Singular Value Normalization and a SpectralNormWrapper. PiperOrigin-RevId: 260938588 06 September 2019, 13:19:11 UTC
1551963 [Automated] Update package documentation. PiperOrigin-RevId: 260133549 06 September 2019, 13:19:01 UTC
e508d28 Update package documentation. PiperOrigin-RevId: 259752487 06 September 2019, 13:18:50 UTC
530ff29 Adding automatic mixed precision support - MNIST MLP example PR: deepmind/sonnet#137 PiperOrigin-RevId: 259734577 06 September 2019, 13:18:40 UTC
9d93f01 Add note regarding tf.switch_case. PiperOrigin-RevId: 259700395 06 September 2019, 13:18:31 UTC
9eb52f4 [Automated] Update package documentation. PiperOrigin-RevId: 258938797 06 September 2019, 13:18:20 UTC
8adaa48 [Automated] Update package documentation. PiperOrigin-RevId: 258930422 06 September 2019, 13:18:09 UTC
55cc9fe Add default getter option to Sonnet custom_getters.Context. PiperOrigin-RevId: 258753325 06 September 2019, 13:18:00 UTC
ccc5779 Strip base_dtype from Sonnet 2. base_dtype only differentiates between reference and non-reference dtypes. In TensorFlow 2 there is no way to produce a reference dtype. PiperOrigin-RevId: 258430193 06 September 2019, 13:17:49 UTC
98258e8 Minor fix in _get_pass_through_gradients_op doc-string. PiperOrigin-RevId: 256973628 06 September 2019, 13:17:37 UTC
b3ff308 Sonnet version update produced on Friday, 05. July 2019 PiperOrigin-RevId: 256668684 05 July 2019, 14:44:57 UTC
e697acd Fix a race condition in MovingAverage. PiperOrigin-RevId: 255576990 05 July 2019, 13:38:23 UTC
5d4684e [Automated] Update package documentation. PiperOrigin-RevId: 255566173 05 July 2019, 13:38:04 UTC
1246c2d Fix typo in doc string. PiperOrigin-RevId: 255156613 05 July 2019, 13:37:55 UTC
a0c9766 Fix TSAN for MovingAverage module tests. PiperOrigin-RevId: 254975013 05 July 2019, 13:37:45 UTC
3a6d29b Allow BatchNormV2 to deal with unknown spatial dimensions. PiperOrigin-RevId: 254963706 05 July 2019, 13:37:35 UTC
92add52 [Automated] Update package documentation. PiperOrigin-RevId: 254927093 05 July 2019, 13:37:24 UTC
a2dd03f Fix typo in error message. PiperOrigin-RevId: 254545767 05 July 2019, 13:36:25 UTC
2efa8e9 Add implementation of GECO: https://arxiv.org/pdf/1810.00597.pdf. This change includes constrained optimization tools enabling users to easily add inequality constraints to their graphs. PiperOrigin-RevId: 254419058 05 July 2019, 13:36:17 UTC
74bbfc4 [Automated] Update package documentation. PiperOrigin-RevId: 254375468 05 July 2019, 13:36:07 UTC
9e81158 [Automated] Update package documentation. PiperOrigin-RevId: 253235248 05 July 2019, 13:35:57 UTC
c1d28b8 Enable ConvND modules to use float64 tensors. PiperOrigin-RevId: 252046103 05 July 2019, 13:35:40 UTC
f6577ed [Automated] Update package documentation. PiperOrigin-RevId: 252013031 05 July 2019, 13:35:30 UTC
5df123e Improve some CPU Conv2D error messages. PiperOrigin-RevId: 251936228 05 July 2019, 13:35:20 UTC
c1565e5 Include absl logging in util dependencies PiperOrigin-RevId: 251654348 05 July 2019, 13:35:11 UTC
df3b144 Switch to absl.logging for all logging functionalities in modules/util.py PiperOrigin-RevId: 251642049 05 July 2019, 13:35:01 UTC
bd8f6bd Generate Add for strings and AddV2 for other types. "Add" implements elements-wise concatenation for string tensors, which makes it non-commutative. PiperOrigin-RevId: 251541420 05 July 2019, 13:34:44 UTC
56c917e Sonnet version update produced on Tuesday, 21. May 2019 PiperOrigin-RevId: 249841355 04 June 2019, 16:38:07 UTC
70a0a26 Update requirements to latest stable versions of TF & TFP PiperOrigin-RevId: 249826100 04 June 2019, 16:37:56 UTC
41234b9 Add citation instructions to Readme. PiperOrigin-RevId: 249824883 04 June 2019, 16:37:45 UTC
4daed34 Update docs to fix rendering in output PiperOrigin-RevId: 249803854 04 June 2019, 16:37:33 UTC
ff2d1ed Use Sonnet logo as favicon. PiperOrigin-RevId: 249798947 04 June 2019, 16:37:22 UTC
2d3194e Version update. - Update documentation - Simplify setup.py and retire dm-sonnet-gpu - Use tf_inspect instead of inspect in sonnet.utils. PiperOrigin-RevId: 246120109 21 May 2019, 15:52:14 UTC
00612ca Bump version to 1.32 and add final changes to changelog. This also adds an __init__ to the protos directory to make bazel pick it up. PiperOrigin-RevId: 243257844 12 April 2019, 19:03:19 UTC
1ff44fd Expose Sonnet 2 via `import sonnet.v2 as snt`. PiperOrigin-RevId: 243101231 12 April 2019, 19:03:09 UTC
f8552f4 Allow named argument 'outputs' in calls to Sonnet modules. PiperOrigin-RevId: 243062408 11 April 2019, 17:49:21 UTC
bd56b26 Python 3 all the things. PiperOrigin-RevId: 243021127 11 April 2019, 17:49:10 UTC
54a9a10 Remove experimental references from documentation. PiperOrigin-RevId: 242887298 11 April 2019, 17:48:56 UTC
63f3680 Updated changelog for revised 1.32 release. PiperOrigin-RevId: 242666638 09 April 2019, 15:01:46 UTC
74e96ec internal change PiperOrigin-RevId: 242629869 09 April 2019, 14:57:48 UTC
ed3f4f6 internal change PiperOrigin-RevId: 242093580 09 April 2019, 14:57:38 UTC
961d7a7 internal change PiperOrigin-RevId: 241523718 09 April 2019, 14:57:28 UTC
2380d9c Add deprecation warning for broken layer norm in ConvLSTM Currently it has a reshape before normalizing so scale and offset are created for product(spatial_dims) * channels instead of just channels PiperOrigin-RevId: 241343128 09 April 2019, 14:57:17 UTC
12f10ed internal change PiperOrigin-RevId: 240950108 09 April 2019, 14:56:59 UTC
90dcf83 Sonnet version update produced on Wednesday, 27. March 2019 PiperOrigin-RevId: 240752885 09 April 2019, 14:56:49 UTC
843ed0a Internal refactor PiperOrigin-RevId: 240314998 27 March 2019, 20:51:18 UTC
c2b3eda Sonnet version update produced on Tuesday, 12. February 2019 PiperOrigin-RevId: 233617272 26 February 2019, 17:34:24 UTC
9c68120 Add mkdocs yaml file. PiperOrigin-RevId: 233616884 26 February 2019, 17:34:15 UTC
bcec0d4 internal change PiperOrigin-RevId: 233565498 12 February 2019, 16:51:22 UTC
b12c680 Support lower precision inputs. PiperOrigin-RevId: 233081564 12 February 2019, 16:51:14 UTC
84e800f internal change PiperOrigin-RevId: 233030785 12 February 2019, 16:51:04 UTC
d11f9c4 A class to linear transform the concatenation of a list of Tensors. It ensures the relative importance of all inputs are similar even if they have very different sizes. PiperOrigin-RevId: 232509624 12 February 2019, 16:50:56 UTC
62b0f6e internal change PiperOrigin-RevId: 232439572 12 February 2019, 16:50:45 UTC
03d9aef Update protobuf dependency. PiperOrigin-RevId: 232317599 04 February 2019, 18:14:37 UTC
381b630 Sonnet version update produced on Tuesday, 29. January 2019 PiperOrigin-RevId: 231442829 04 February 2019, 18:14:29 UTC
e1f8283 internal change PiperOrigin-RevId: 231365216 29 January 2019, 17:35:41 UTC
8b4942f Internal change PiperOrigin-RevId: 231186348 29 January 2019, 17:35:34 UTC
ae7e001 Internal change. PiperOrigin-RevId: 231069857 29 January 2019, 17:35:28 UTC
674007e Make module/connection stacks thread local. PiperOrigin-RevId: 231065224 29 January 2019, 17:35:21 UTC
f138492 Internal Change. PiperOrigin-RevId: 230959989 29 January 2019, 17:35:15 UTC
3a9e904 internal change PiperOrigin-RevId: 230881819 29 January 2019, 17:35:08 UTC
eb46086 Allow control over max pondering steps. PiperOrigin-RevId: 230704013 29 January 2019, 17:35:02 UTC
f451e9a Add hyperlinks to documentation. PiperOrigin-RevId: 230691280 29 January 2019, 17:34:56 UTC
e478c60 Fix headers in installation instructions. PiperOrigin-RevId: 230691260 29 January 2019, 17:34:49 UTC
11e00f7 Remove Pandoc and Sphinx configuration files. PiperOrigin-RevId: 230690314 29 January 2019, 17:34:43 UTC
6c53258 internal change PiperOrigin-RevId: 230302480 29 January 2019, 17:34:36 UTC
249c817 internal change PiperOrigin-RevId: 228307386 29 January 2019, 17:34:29 UTC
675e849 Increase test size to accommodate sanitizers PiperOrigin-RevId: 228141037 29 January 2019, 17:34:19 UTC
2f7a52f Remove `@snt.experimental.reuse_vars`. Please use `@snt.reuse_variables` instead! PiperOrigin-RevId: 228063197 29 January 2019, 17:34:12 UTC
a9c6a8d Allowing Conv2D to accept unicode, in addition to str. PiperOrigin-RevId: 227748726 29 January 2019, 17:34:06 UTC
0bfb404 Add MLP MNIST example. PiperOrigin-RevId: 227704815 29 January 2019, 17:34:00 UTC
e95ba5c Add wrapt to required packages. Fixes https://github.com/deepmind/sonnet/issues/115 PiperOrigin-RevId: 227703904 29 January 2019, 17:33:53 UTC
a12ce4d internal change PiperOrigin-RevId: 227485581 29 January 2019, 17:33:46 UTC
bbdf2ae Put reused variables in _all_variables when using nested modules. It was possible before to end up in an inconsistent state if inside `_build` `ParentModule` assigned `self.child = Child()` and in your training loop you used `mod.child.variables` (or `get_all_variables()`): ``` mod = ParentModule() for record in inputs: with tf.GradientTape() as tape: outs = mod(record) vars = mod.child.variables # After first iteration this would be empty. grads = tape.gradient(outs, vars) ``` While I suspect this sort of thing would be more common in eager mode, the bug still exists in graph mode. PiperOrigin-RevId: 227178237 29 January 2019, 17:33:40 UTC
91bab7b internal change PiperOrigin-RevId: 227108833 29 January 2019, 17:33:33 UTC
fcf20ed Enable Python 3 in Sonnet's py_binary rules. PiperOrigin-RevId: 227011720 29 January 2019, 17:33:27 UTC
e0988b1 Remove some dependencies on internal TensorFlow symbols. PiperOrigin-RevId: 226960072 29 January 2019, 17:33:20 UTC
0400b3d internal change PiperOrigin-RevId: 226476878 29 January 2019, 17:33:14 UTC
fe13d13 internal change PiperOrigin-RevId: 226179552 29 January 2019, 17:33:03 UTC
79d920c Sonnet version update produced on Wednesday, 19. December 2018 PiperOrigin-RevId: 226148846 29 January 2019, 17:32:56 UTC
bb2e382 Update changelog for version 1.28 PiperOrigin-RevId: 226008459 29 January 2019, 17:32:49 UTC
39e817b FIX: setup.py.tmpl referenced wrong package name setup.py.tmpl referenced "tensor-probability-gpu" instead of "tensorflow-probability-gpu". this caused installation via pip to fail since dm-sonnet-gpu==1.25 GIT_ORIGIN_REV_ID=b18ca77dd73795d5c18dbf2a1e895471493c15f2 PiperOrigin-RevId: 226004212 29 January 2019, 17:32:43 UTC
5995647 Allow for kwargs to be forwarded to sub-modules in DeepRNN. Useful for propagating 'is_training'. PiperOrigin-RevId: 225976178 29 January 2019, 17:32:32 UTC
8b557d8 internal change PiperOrigin-RevId: 225181012 29 January 2019, 17:32:25 UTC
a35746d ConvNet2D{Transpose}: Filter the kwargs passed to the normalizer. This makes it possible to connect the module with is_training=True, even if the normalization module selected is something which does not support that (eg LayerNorm). Normal practice would be to call the module with is_training inside a try block, catch any errors and then reconnect without is_training. However, that will generally have created some variables internally, so global (tf.Graph-level) state has been changed, producing errors. This solution works for any normalization module which has a signature that does not contain **kwargs, because that makes it impossible to introspect over whether the flag is supported or not. PiperOrigin-RevId: 224994092 29 January 2019, 17:32:14 UTC
0c8a946 1. Fix the sequence reading order for backward unroll. The order should be decreasing but was increasing. 2. Fix RNN unroll by feeding correct state (the previous state) to each unroll step. The state was always the initial state. PiperOrigin-RevId: 224174351 29 January 2019, 17:32:07 UTC
a64805b Support non-2D recurrent state in pondering RNN. (Still relies on leading dimension being the batch dimension.) PiperOrigin-RevId: 223522357 29 January 2019, 17:32:01 UTC
back to top