deddd25 | Frank Seide | 30 May 2017, 21:56:43 UTC | fixed the Communicator use, which is not germaine to this tutorial (although teh bug is really in Communicator itself, which should be a dummy in non-1Bit-SGD builds) | 30 May 2017, 21:56:43 UTC |
55ad83c | Project Philly | 30 May 2017, 16:19:06 UTC | Integrate t-ivrodr/fix_for_bn_examples into master | 30 May 2017, 16:19:06 UTC |
6f3f3f0 | Project Philly | 30 May 2017, 15:16:57 UTC | Integrate wilrich/fixMessageBug into master | 30 May 2017, 15:16:57 UTC |
90a5c1e | Project Philly | 30 May 2017, 14:07:54 UTC | Integrate mahilleb/testTolerance into master | 30 May 2017, 14:07:54 UTC |
5ae4f23 | Willi Richert | 30 May 2017, 13:43:23 UTC | Fix message formatting bug; add test | 30 May 2017, 14:00:20 UTC |
698dc3e | Willi Richert | 30 May 2017, 13:23:07 UTC | autopep | 30 May 2017, 13:23:07 UTC |
6a8cdf7 | Project Philly | 30 May 2017, 13:16:41 UTC | Integrate eldak/addingDeprecatedMessage into master | 30 May 2017, 13:16:41 UTC |
c0ed1cf | Ivan Rodriguez | 29 May 2017, 10:52:22 UTC | Restoring timeout | 30 May 2017, 13:10:43 UTC |
fe6c089 | Mark Hillebrand | 30 May 2017, 11:21:10 UTC | Tests/EndToEndTests/CNTKv2Python/Examples/BN_Inception_CIFAR10_test.py: bump tolerance | 30 May 2017, 11:36:17 UTC |
6144630 | Ivan Rodriguez | 29 May 2017, 10:52:22 UTC | Addressing comments in CR | 30 May 2017, 10:33:24 UTC |
b9bedb0 | Ivan Rodriguez | 29 May 2017, 10:52:22 UTC | Adding true division for BN examples and progress printer | 30 May 2017, 09:54:06 UTC |
12df59d | Eldar Akchurin | 29 May 2017, 14:17:31 UTC | Adding deprecation message | 30 May 2017, 09:33:35 UTC |
67545b8 | Frank Seide | 30 May 2017, 02:08:07 UTC | more CR feedback; renamed back to 99, this time with git mv as requested, then re-added to the VS Solution | 30 May 2017, 02:08:07 UTC |
22ff607 | Frank Seide | 30 May 2017, 01:48:05 UTC | more feedback; temporarily renamed back to 99 since CodeFLow cannot handle this | 30 May 2017, 01:48:05 UTC |
befd9ac | Frank Seide | 29 May 2017, 23:48:32 UTC | test for 050 now works; finalized the notebook | 29 May 2017, 23:48:32 UTC |
91f4a74 | Frank Seide | 29 May 2017, 23:10:28 UTC | addressed CR feedback; added test for 050 tutorial | 29 May 2017, 23:10:28 UTC |
e627554 | Frank Seide | 29 May 2017, 21:10:31 UTC | added 050 tutorial to CNTK.sln after fixing CNTK.sln | 29 May 2017, 21:10:31 UTC |
97096e3 | Frank Seide | 29 May 2017, 21:06:58 UTC | finally got cntk.sln | 29 May 2017, 21:06:58 UTC |
56dd32f | Frank Seide | 29 May 2017, 20:51:06 UTC | updated CNTK.sln | 29 May 2017, 20:51:06 UTC |
498031b | Frank Seide | 29 May 2017, 20:47:56 UTC | merged CNTK.sln back from master | 29 May 2017, 20:47:56 UTC |
1e948b0 | Frank Seide | 29 May 2017, 20:45:57 UTC | Merge branch 'master' of https://github.com/Microsoft/cntk into fseide/99 | 29 May 2017, 20:45:57 UTC |
70328b9 | Frank Seide | 29 May 2017, 20:44:37 UTC | renamed 099 to 050 | 29 May 2017, 20:44:37 UTC |
06aaf70 | Project Philly | 29 May 2017, 18:59:45 UTC | Integrate alrezni/restore_dict into master | 29 May 2017, 18:59:45 UTC |
bef84a0 | Alexey Reznichenko | 24 May 2017, 12:59:35 UTC | Fix Trainer.restore_from_checkpoint Make sure that it returns a python dictionary containing previously saved persistent state. | 29 May 2017, 18:32:16 UTC |
e3c6f47 | Alexey Reznichenko | 22 May 2017, 16:18:22 UTC | Use Optional<bool> for MBSourceConfig isMultithreaded | 29 May 2017, 18:26:28 UTC |
8505c6e | Project Philly | 29 May 2017, 15:28:52 UTC | Integrate wilrich/fixUserMBSource into master | 29 May 2017, 15:28:52 UTC |
7b8556a | Mark Hillebrand | 29 May 2017, 12:39:54 UTC | Source/Math/CuDnnBatchNormalization.cu: address CR comment | 29 May 2017, 12:39:54 UTC |
11bd2ca | Willi Richert | 24 May 2017, 19:16:10 UTC | Add default checkpointing to UserMinibatchsource | 29 May 2017, 09:47:54 UTC |
973d29b | Frank Seide | 29 May 2017, 08:49:58 UTC | Merge branch 'master' of https://github.com/Microsoft/cntk into fseide/99 | 29 May 2017, 08:49:58 UTC |
4e7251a | Frank Seide | 29 May 2017, 08:49:24 UTC | forgot a @staticmethod decorator; somehow CNTK.sln did not merge | 29 May 2017, 08:49:24 UTC |
42bdf5a | Frank Seide | 29 May 2017, 03:41:36 UTC | Merge branch 'master' of https://github.com/Microsoft/cntk into fseide/99 | 29 May 2017, 03:41:36 UTC |
867c847 | Frank Seide | 29 May 2017, 03:39:56 UTC | did a full run of tutorial 099 | 29 May 2017, 03:39:56 UTC |
ed33d18 | Frank Seide | 29 May 2017, 03:26:54 UTC | completed 099 turorial | 29 May 2017, 03:26:54 UTC |
438e165 | Frank Seide | 29 May 2017, 01:13:53 UTC | io now imports transformx; updated 099 tutorial | 29 May 2017, 01:13:53 UTC |
494ee89 | KeDengMS | 28 May 2017, 23:01:52 UTC | Fix some docstrings in crosstalk and optimized_rnnstack | 28 May 2017, 23:01:52 UTC |
7aca401 | Frank Seide | 28 May 2017, 22:38:07 UTC | updates to 099 | 28 May 2017, 22:38:07 UTC |
f2873b0 | Frank Seide | 28 May 2017, 21:55:46 UTC | edits to 099 tutorial; renamed cntk.FunctionOf to cntk.Function.with_signature, since with the mandatory cntk. prefix, the shorter name does not add to lightness, and the longer form is clearer; new function utils.test_utils.set_device_from_pytest_env() so we can simplify the Jupyter notebooks; fixed a spelling error in the HTK example; some comment edits to 1stSteps examples | 28 May 2017, 21:55:46 UTC |
7fe5045 | Frank Seide | 28 May 2017, 21:26:07 UTC | updated CNTK_099 tutorial | 28 May 2017, 21:26:07 UTC |
c1c55cf | Frank Seide | 28 May 2017, 21:08:50 UTC | updated CNTK_099 tutorial | 28 May 2017, 21:08:50 UTC |
a24b971 | Frank Seide | 28 May 2017, 21:08:02 UTC | added CNTK_099 tutorial | 28 May 2017, 21:08:02 UTC |
71b8a35 | Amit Agarwal | 26 May 2017, 08:30:39 UTC | CNTK v2 library: Add Binary convolution native user-defined Function library. | 28 May 2017, 05:52:08 UTC |
ffeca8f | Amit Agarwal | 25 May 2017, 20:30:21 UTC | CNTK v2 library: Fix a bug in propagation of free dimensions originating at user-defined Functions. | 28 May 2017, 05:52:08 UTC |
324140b | Project Philly | 28 May 2017, 02:20:51 UTC | Integrate chenta/batch_universal_learner into master | 28 May 2017, 02:20:51 UTC |
faa2c05 | Project Philly | 27 May 2017, 23:58:29 UTC | Integrate fseide/fixes into master | 27 May 2017, 23:58:29 UTC |
06af823 | Cheng Tang | 27 May 2017, 00:15:06 UTC | fix comments | 27 May 2017, 23:31:43 UTC |
017e594 | Cheng Tang | 26 May 2017, 23:17:01 UTC | update interface name. update python api to access lambda with parameters and gradients | 27 May 2017, 23:31:43 UTC |
703d482 | Cheng Tang | 26 May 2017, 17:45:18 UTC | fix csharp binding failure | 27 May 2017, 23:31:43 UTC |
63e34be | Cheng Tang | 26 May 2017, 06:55:23 UTC | integrate with universal learner | 27 May 2017, 23:31:43 UTC |
9488c88 | Cheng Tang | 26 May 2017, 00:01:51 UTC | batch universal learner | 27 May 2017, 23:31:42 UTC |
c48b4de | Project Philly | 27 May 2017, 20:54:39 UTC | Integrate nikosk/assign_fix into master | 27 May 2017, 20:54:39 UTC |
9215b5e | Frank Seide | 27 May 2017, 03:54:10 UTC | two more fixes along the same line as the previous commit | 27 May 2017, 03:54:10 UTC |
ccd975f | Nikos Karampatziakis | 27 May 2017, 02:57:06 UTC | Merge remote-tracking branch 'origin/master' into nikosk/assign_fix | 27 May 2017, 02:57:06 UTC |
482c259 | Nikos Karampatziakis | 27 May 2017, 01:58:11 UTC | make doctest more robust | 27 May 2017, 01:58:11 UTC |
8850632 | Frank Seide | 27 May 2017, 01:31:09 UTC | spelling error | 27 May 2017, 01:31:09 UTC |
1e27a6c | Frank Seide | 27 May 2017, 01:16:13 UTC | Merge branch 'master' of https://github.com/Microsoft/cntk into fseide/fixes | 27 May 2017, 01:16:13 UTC |
6282f80 | Frank Seide | 27 May 2017, 00:59:39 UTC | fixed test failures | 27 May 2017, 00:59:39 UTC |
6507207 | Nikos Karampatziakis | 27 May 2017, 00:58:53 UTC | incorporate cr feedback | 27 May 2017, 00:58:53 UTC |
d001107 | Frank Seide | 27 May 2017, 00:51:23 UTC | added test for MNIST example; added to doc of Dropout(); added a fallback for downloading the MNIST data in case sklearn is not available | 27 May 2017, 00:51:23 UTC |
32605e3 | Nikos Karampatziakis | 27 May 2017, 00:19:41 UTC | round gumbel_like one digit more | 27 May 2017, 00:19:41 UTC |
f8a452c | Frank Seide | 26 May 2017, 23:30:54 UTC | changed MaxPooling to (2,2) to make it deterministic (for our testing process) | 26 May 2017, 23:30:54 UTC |
d048378 | Nikos Karampatziakis | 26 May 2017, 23:30:42 UTC | Merge remote-tracking branch 'origin/master' into nikosk/random_ops2 | 26 May 2017, 23:30:42 UTC |
941cf5e | Nikos Karampatziakis | 26 May 2017, 23:29:48 UTC | incorporate cr feedback | 26 May 2017, 23:29:48 UTC |
c79a741 | Frank Seide | 26 May 2017, 22:49:37 UTC | addressed CR feedback | 26 May 2017, 22:49:37 UTC |
d332ae2 | Frank Seide | 26 May 2017, 22:31:28 UTC | addressed CR feedback | 26 May 2017, 22:31:28 UTC |
d271bba | Frank Seide | 26 May 2017, 22:11:51 UTC | MNIST example increased the LR cut threshold | 26 May 2017, 22:11:51 UTC |
6e30cf5 | Nikos Karampatziakis | 26 May 2017, 19:00:13 UTC | bump timestamp of assign ref in v2 | 26 May 2017, 19:00:13 UTC |
e9b5422 | Zhou Wang | 23 May 2017, 21:39:24 UTC | ignore _wcsdup on Linux | 26 May 2017, 14:38:22 UTC |
2cd76de | Frank Seide | 26 May 2017, 03:45:52 UTC | (edited a message) | 26 May 2017, 03:45:52 UTC |
8c5c6d0 | Frank Seide | 26 May 2017, 03:44:07 UTC | MinibatchSourceFromData now implements number_of_workers > 1; MNIST example implements data-parallel training; disabled printing the built-info, because there is a separate call for it that the user can call explicitly if needed | 26 May 2017, 03:44:07 UTC |
48fd5ef | Frank Seide | 26 May 2017, 01:26:23 UTC | undid learning_rate as a @property, since the distributed learners have no Python wrappers where I could make the same change (or do they?); bug fix in train()'s empty test | 26 May 2017, 01:26:23 UTC |
d74eacd | Frank Seide | 26 May 2017, 01:02:09 UTC | updated graph_test after master merge to __str__() again | 26 May 2017, 01:02:09 UTC |
da31323 | Chris Basoglu | 26 May 2017, 00:51:28 UTC | Update ConvNet_CIFAR10_DataAug_Distributed.py | 26 May 2017, 00:51:28 UTC |
33820d2 | Chris Basoglu | 26 May 2017, 00:48:59 UTC | Update ConvNet_CIFAR10_DataAug_Distributed.py | 26 May 2017, 00:48:59 UTC |
16cd872 | Frank Seide | 26 May 2017, 00:43:56 UTC | bug fix: train() should return empty statistics in case no training was executed (fall-through from checkpoint); added cross-validation to MNIST sample | 26 May 2017, 00:43:56 UTC |
745b492 | Frank Seide | 25 May 2017, 22:13:31 UTC | cleaned up CheckPointConfig and more | 25 May 2017, 22:13:31 UTC |
8f90fdc | Project Philly | 25 May 2017, 22:11:49 UTC | Integrate chenta/align_axis into master | 25 May 2017, 22:11:49 UTC |
562667d | Frank Seide | 25 May 2017, 22:06:29 UTC | fixed a test after merge | 25 May 2017, 22:06:29 UTC |
b097a90 | Frank Seide | 25 May 2017, 21:55:04 UTC | addressed CR feedback & Jenkins errors; changed learner.learning_rate to a @property | 25 May 2017, 21:55:04 UTC |
a66502d | Nikos Karampatziakis | 25 May 2017, 21:10:46 UTC | incorporate CR feedback | 25 May 2017, 21:10:46 UTC |
b03c007 | Frank Seide | 25 May 2017, 20:52:06 UTC | minor fixes after master merge | 25 May 2017, 20:52:06 UTC |
8acef57 | Frank Seide | 25 May 2017, 20:46:15 UTC | merged from master | 25 May 2017, 20:46:15 UTC |
410be7d | Frank Seide | 25 May 2017, 19:13:19 UTC | added MNIST example for complex training; added checkpointing to MNIST example, but not working; added get_checkpoint_state() and restore_from_checkpoint() to UserMinibatchSource as NotImplementedError; added checkpointing to MinibatchSourceFromData (but not working) | 25 May 2017, 19:13:19 UTC |
d415e73 | Frank Seide | 25 May 2017, 18:50:50 UTC | train() now understands TestConfig (but only one test at the end); added arg map and criterion to TestConfig; MNIST example now using TestConfig to test | 25 May 2017, 18:50:50 UTC |
e981e3a | Frank Seide | 25 May 2017, 18:22:06 UTC | replaced the various callback parameters in train() and test() with a single callbacks list | 25 May 2017, 18:22:06 UTC |
bb2c3d9 | Cheng Tang | 25 May 2017, 17:54:00 UTC | fix a bug during get item | 25 May 2017, 17:54:00 UTC |
a6c7ec6 | Frank Seide | 25 May 2017, 17:01:49 UTC | train() and test() now return stats not as tuples but as an object; MNIST example now uses a proper convolutional model from the ConvNet example | 25 May 2017, 17:01:49 UTC |
4d994b3 | Frank Seide | 25 May 2017, 05:08:18 UTC | changed LogisticRegression examples to not use csr_matrix.argmax since that is only available in scipy 19+; added MNIST.py first cut | 25 May 2017, 05:08:18 UTC |
4110f6f | REDMOND\sayanpa | 25 May 2017, 02:34:36 UTC | Merge branch 'master' of https://github.com/Microsoft/CNTK into sayanpa/tutfix | 25 May 2017, 02:34:36 UTC |
c14c7ca | REDMOND\sayanpa | 25 May 2017, 02:29:53 UTC | Change data source for the finance tutorial | 25 May 2017, 02:29:53 UTC |
b7e51fa | Cheng Tang | 24 May 2017, 18:35:22 UTC | fix comments in cr | 25 May 2017, 01:30:03 UTC |
263e19c | Cheng Tang | 23 May 2017, 17:28:35 UTC | add python global flag and align axis when call __getitem__ in keras mode | 25 May 2017, 01:30:03 UTC |
92ebf2b | Frank Seide | 25 May 2017, 01:25:19 UTC | fixed tests after adding __str__() and Value.one_hot() was fixed to accept non-sequences | 25 May 2017, 01:25:19 UTC |
06b0060 | Frank Seide | 24 May 2017, 22:52:01 UTC | addressed CR feedback | 24 May 2017, 22:52:01 UTC |
c866e56 | REDMOND\sayanpa | 24 May 2017, 21:43:12 UTC | Merge branch 'master' of https://github.com/Microsoft/CNTK into sayanpa/tutfix | 24 May 2017, 21:43:12 UTC |
ddf8f53 | Emad Barsoum | 24 May 2017, 18:55:44 UTC | Minor cleanup. | 24 May 2017, 18:55:44 UTC |
d2f83e9 | Emad Barsoum | 24 May 2017, 18:23:32 UTC | Merge branch 'master' into ebarsoum/fixes For checkin... | 24 May 2017, 18:23:32 UTC |
be119fb | Emad Barsoum | 24 May 2017, 18:21:56 UTC | Fix assign node to mark the destination node as dirty. | 24 May 2017, 18:21:56 UTC |
dae6ff8 | Frank Seide | 24 May 2017, 18:01:43 UTC | (minor CR feedback) | 24 May 2017, 18:01:43 UTC |
51ec49c | Frank Seide | 24 May 2017, 17:53:25 UTC | addressed most namespaces | 24 May 2017, 17:53:25 UTC |
aac0ec4 | Frank Seide | 24 May 2017, 17:40:29 UTC | added defaults to train(), minibatch_size=32 and max_epochs=1, which both follow defaults familiar from other toolkits; added 1stSteps examples to samples.json | 24 May 2017, 17:40:29 UTC |