a1e10ff9e7
Bump version numbers
2017-04-21 15:58:00 +02:00
3542533708
Py 3.6 build & test
2017-04-13 14:19:19 +02:00
100ae50494
Adding tests to layers library, removing previous non-layers constructors and updating distributed seq2seq.
2017-04-10 13:51:30 +02:00
94993f3c81
Integrate alrezni/v2_dropout into master
2017-04-04 09:16:31 -07:00
859296557d
Add local worker's state checkpointing
...
Add a mechanism to preserve the local state of a distributed worker
(together with the external state aggregated by the main node and saved
inside the checkpoint).
2017-04-04 16:07:48 +02:00
861bfbd714
Bump versions
2017-04-03 18:16:31 +02:00
8a9020f78e
CNTK v2 library: Migrate past_value and future_value to sequence
2017-03-30 12:11:20 -07:00
0d2879eab5
CNTK v2 library: Remove remnants of the use of deprecated input_variable and placeholder_variable methods
2017-03-30 12:11:20 -07:00
d251977f38
Merge with master.
2017-03-29 15:05:51 -07:00
05147afa9e
More fixes and examples for Parameter and Constant.
2017-03-29 14:52:37 -07:00
43cd200cf0
Fix comments and remaining notebooks.
2017-03-29 10:41:59 -07:00
9adade3311
Adapting examples; set MinibatchData.data to be a property
2017-03-29 17:09:13 +02:00
90287b6db7
Fix Notebook...
2017-03-29 02:07:15 -07:00
2a05f2157f
Integrate fseide/s2spy into master
2017-03-28 16:17:34 -07:00
ccf1c280ce
Updated the tx learning tutorial test settings in jupyter
2017-03-28 13:57:56 -07:00
10080aa64b
Reduced runtime in Fast mode and fixed an indexing bug with current breaking change in master
2017-03-28 11:50:50 -07:00
3843051636
Merge branch 'master' of https://github.com/Microsoft/CNTK into sayanpa/txlrn
2017-03-28 10:27:44 -07:00
69be0723a5
Merge branch 'master' of https://github.com/Microsoft/cntk into fseide/s2spy
2017-03-28 08:55:10 -07:00
f45edca575
bug fix in UnfoldFrom() API: initial_state must be a parameter;
...
removed unused left-over parameter map_state_function from UnfoldFrom()
2017-03-28 08:38:50 -07:00
503983b0d1
Add an ability to limit the number of sweeps for training
...
* Add MinibatchSourceConfig struct that makes construction of a reader
a little more straight-forward (esp. on the c++ side)
* Refactor python implementation of MinibatchSource.
* Add support for specifying the maximum number of sweeps over the input
dataset when instating a minibatch source.
2017-03-28 13:50:17 +02:00
97782a2098
Incorporated CR feedback
2017-03-27 19:13:32 -07:00
156028a4dc
Updated to latest API changes
2017-03-27 15:11:56 -07:00
e602908271
Merge branch 'master' of https://github.com/Microsoft/CNTK into sayanpa/txlrn
2017-03-26 11:27:41 -07:00
2bd14b7026
CNTK v2 library: Several UserDefinedFunction features and bug fixes
...
1) Fix for Premature UDF backprop state destruction in python SWIG wrapper
(Willi and Eldar)
2) Add ability for output variable to specify that gradients are not
needed
3) Add support for letting backward implementation of a UDF be a no-op which
results in no gradients being propagated to the inputs for that UDF node
4) Change UDF bckward signature to be passed a map for filling input
gradients whevener the UDF has multiple inputs
5) Performance fix to call the UDF backward methid just once for all
inputs needing gradients instead of separate calls for each input
2017-03-25 21:19:29 -07:00
5d79f54d15
Merge branch 'master' of https://github.com/Microsoft/CNTK into sayanpa/txlrn
2017-03-24 19:41:05 -07:00
138ce5299e
Fix break after merge with master.
2017-03-24 15:16:02 -07:00
7316050cca
More fixes.
2017-03-24 14:47:38 -07:00
a956045d2b
More fixes.
2017-03-24 13:53:25 -07:00
1e90099ca8
Fix notebooks.
2017-03-24 11:53:54 -07:00
45ff901bf7
Added a flag to force data download during testing
2017-03-24 09:29:05 -07:00
e1021f702f
Testing forced download of data
2017-03-23 20:27:45 -07:00
6725821e45
Added cached data download for jenkins testing
2017-03-23 18:43:16 -07:00
1d4caaa7df
Merge branch 'master' of https://github.com/Microsoft/CNTK into sayanpa/txlrn
2017-03-23 11:59:18 -07:00
6220171165
Update CNTK_103B_MNIST_FeedForwardNetwork.ipynb
...
modify the text to match with the variable value
2017-03-23 07:33:20 -04:00
a9607caadf
CNTK v2 library: Update all tests and examples to properly use ops.input and sequence.input methods for creating input variables
2017-03-22 06:44:59 -07:00
c4a221d227
update seq2seq tutorial 204 to Layers library version
2017-03-21 13:54:04 +01:00
70507ad1ca
Merged with master and some code cleanup
2017-03-20 17:40:56 -07:00
f0b3a48925
Unicode exception handling in Linux and test threshold tuning
2017-03-20 16:00:34 -07:00
3fe6aa03bc
fixed a print bug
2017-03-20 16:00:34 -07:00
9ca0ad800c
fixed print error
2017-03-20 16:00:34 -07:00
c4325e1fc5
fixed a utf-8 bug fix
2017-03-20 16:00:33 -07:00
47a0141134
Increase number of iterations
2017-03-20 16:00:33 -07:00
c952f1d260
Bug fixes in file loading and exception handling
2017-03-20 16:00:33 -07:00
b1105b53fe
fix ascii encoding for py 3.4
2017-03-20 16:00:33 -07:00
3ed8439536
updated number of iterations and a bug
2017-03-20 16:00:32 -07:00
dd75a79580
fixed py2.7 bug
2017-03-20 16:00:32 -07:00
201d4a2bb9
Fixed a accuracy bug for p2.7
2017-03-20 16:00:32 -07:00
17a6ad9681
Updated test thresholds
2017-03-20 16:00:32 -07:00
444de8c528
Bug fix for py2.7 with urllib.request
2017-03-20 16:00:31 -07:00
0eac3877c4
Updated test paths
2017-03-20 16:00:31 -07:00
f4c1c0c904
Added Transfer Learning tutorial contribution from Mike Lanzetta, modified it and added tests
2017-03-20 16:00:30 -07:00
3eb95f8b8e
Add initial problem motivation for Transfer Learning notebook.
2017-03-20 15:57:02 -07:00
39a892fc17
Minor change to verbiage to clarify initial image from Convolution layers.
2017-03-20 15:57:02 -07:00
e4a2af8ffe
CNTK301-TransferLearning. Finishing up response to review comments for initial review. Removing old notebook name.
2017-03-20 15:57:02 -07:00
9403525e2e
Partially addressing review comments. More to come.
2017-03-20 15:57:01 -07:00
1c23bc8a15
Adding transfer learning notebook
2017-03-20 15:57:01 -07:00
cbedd91ee1
Add CR review for DCGAN
2017-03-18 18:38:23 -07:00
49da9a4060
Added CR feedback
2017-03-17 19:12:16 -07:00
984358de16
Deleted renamed notebook and corresponding test
2017-03-17 16:21:57 -07:00
b831a9a857
Updated notebooks, tests and docs
2017-03-17 16:19:34 -07:00
d60958f7bb
Merge branch 'master' of https://github.com/Microsoft/CNTK into sayanpa/dcgan3
2017-03-17 15:10:08 -07:00
e364289aa1
Updated dcgan test and nb text'
2017-03-17 15:07:25 -07:00
8da12bcd49
Merge with master.
2017-03-17 10:36:40 -07:00
124f07a0cc
Another attempt at streamlining device selection
...
Update device selection API as agreed:
* drop BestDevice() and DefaultDevice()
* add an option to specify excluded devices that should not be used as
globally default (neither auto-selected nor explicitly specified)
* add a mechanism to query device properties
* add a method to determine if a device is locked or not (in real-time)
* in SetDefaultDevice, offer an option to exclusively lock the specified
device.
2017-03-17 12:34:54 +01:00
a31c1bccad
adding dcgan tutorial
2017-03-16 22:53:39 -07:00
72b9858189
Fix distributed training.
2017-03-16 18:31:09 -07:00
4a93dd0021
Fix Parameters class.
2017-03-16 15:41:15 -07:00
622798a4bd
Fix sparce test.
2017-03-16 14:18:42 -07:00
e25dd44c70
Merge with master.
2017-03-15 22:10:17 -07:00
d86c231d1f
Move all examples and notebook to use the new namespace.
2017-03-15 22:07:51 -07:00
97e674c934
Bump versions
2017-03-15 17:16:54 +01:00
4143756b2c
Integrate amtiaga/bugFixesForReasonNet into master
2017-03-14 14:49:16 -07:00
a5c581064f
Integrate f52ab2a into master
2017-03-14 14:23:01 -07:00
a195ebf73a
CNTK v2 library: Fix output Value object shapes to not have the sequence axis if the variable the Value corresponds to does not have a sequence axis
2017-03-14 12:04:48 -07:00
d15fe33db9
import os
2017-03-14 10:05:09 +01:00
f52ab2a03a
Perform selected action at each time step
...
Currently resetting the environment
2017-03-10 14:02:41 -06:00
1bdd3da4e4
reverted depth_first_search distance/depth in stack variable to prior;
...
deleted some unused left-over code;
fixed some import/namespace issues with new tests
2017-03-03 14:02:18 -08:00
42256b68ce
updated 204 tutorial to new signature of LSTM()
2017-03-03 11:50:23 -08:00
51a3ee5acb
merged from master
2017-03-03 10:30:11 -08:00
2dcfab9c97
fixed a syntax error in a tutorial (merge error??)
2017-03-03 08:44:46 -08:00
6256c6e355
Updated examples to use the latest progress writer API
2017-03-02 17:57:57 -08:00
736d2f006c
merged from master
2017-03-02 11:46:22 -08:00
28b749db39
Update CNTK_106B_LSTM_Timeseries_with_IOT_Data.ipynb
2017-03-01 12:05:13 -08:00
e391687e43
simplified layers_test.py (no more fooDict);
...
changed all example @Function with signature to @FunctionWithSignature for python 2.7 happiness;
fixed an incorrect use of default values in word_rnn.py and 207 tutorial;
updated a CS test w.r.t. LanguageUnderstanding.py
2017-02-24 17:31:02 -08:00
ef3aa02038
CNTK v2 library: Fixed a couple of tutorials that were using numpy imnterop to properly specify the dynamic axes of inputs; earlier the models were specifying inputs to have both sequence and dynamic axes though the fed numpy array data did not account for the sequence axis causing the batch dimension to be wrongly interpreted as the sequence dimension. This was working earlier due to a bug in interpreting numpy data which always interpreted the leading axis to be the bacth axis regardless of the actual dynamic axis the input has; fixing that bug exposed the issue with these notebooks
2017-02-23 13:39:36 -08:00
1b4ec4feeb
Merge remote-tracking branch 'origin/master' into mahilleb/PostReleaseMerges
...
Resolved conflict with:
bindings/python/doc/tutorials.rst
2017-02-23 20:47:44 +01:00
dc0dde69ca
Bump versions
...
Conflict resolved for:
bindings/python/doc/tutorials.rst
2017-02-23 17:10:19 +01:00
a05941709f
read csv with np.float32 to avoid warning
2017-02-22 13:12:59 -08:00
1337cddcb0
Merge branch 'master' into guschmue/tut-106b
2017-02-21 13:10:30 -08:00
4affe72568
HTK deserializers and Reader Transforms
...
Squashed commit of the following:
commit f0250727423dcea2fa138e519562fc868b60633e
Author: Nikos Karampatziakis <nikosk@microsoft.com>
Date: Fri Feb 17 16:36:10 2017 -0800
fix a typo
commit 87d4a2119fc0011bb3199c9dcefead913a92b358
Author: Nikos Karampatziakis <nikosk@microsoft.com>
Date: Fri Feb 17 16:23:20 2017 -0800
Moved C++ test to use HTK deserializers
commit 7b52fe1e5de282787e3e59d1a798336976f3d331
Author: Nikos Karampatziakis <nikosk@microsoft.com>
Date: Fri Feb 17 16:07:23 2017 -0800
final api for htk deserializers
commit e21495f3db074e2fa95ac30933925ed30262a7ba
Author: Nikos Karampatziakis <nikosk@microsoft.com>
Date: Fri Feb 17 15:19:20 2017 -0800
test
commit e3951291d060acad2caf49e3a4d6781da354835a
Author: Nikos Karampatziakis <nikosk@microsoft.com>
Date: Fri Feb 17 14:45:06 2017 -0800
fix
commit 3096607bd2fcbaa93e970daccdb503e29ac4d37c
Author: Nikos Karampatziakis <nikosk@microsoft.com>
Date: Fri Feb 17 14:27:27 2017 -0800
Incorporated CR feedback.
commit 2e3c9d31fa1e4dc141fbc6f2ca3ce69d1b5ad855
Merge: 3e116ed e70eac3721a6f70bf2f460bf2f4615737efhttps://github.com/MSharman/CNTK into mudithtk
commit d9b5fdf0ebab2ff4175d7ae27d1e3cc52d1c195d
Author: Nikos Karampatziakis <nikosk@microsoft.com>
Date: Thu Feb 16 10:01:46 2017 -0800
Various fixes
commit 45d149ef1013fc118de142d96c19110753331aab
Author: Nikos Karampatziakis <nikosk@microsoft.com>
Date: Sat Feb 11 17:16:47 2017 -0800
added docs for HTK deserializers
commit 3baaaf019dd95990f2f8962e35b9bbd0f37f768b
Author: Nikos Karampatziakis <nikosk@microsoft.com>
Date: Sat Feb 11 17:00:54 2017 -0800
update htk test to conform to new trainer
commit 9a39a41be1d2d553f9730dde1912eecbe0805dfb
Merge: 05d211b fd2e796b15978174fa7f415737ef843
2017-02-17 16:39:58 -08:00
721a6f71a2
Integrate 2eaef84d7b into master
2017-02-16 18:55:26 -08:00
a8eeac4123
merge with master
2017-02-16 18:15:10 -08:00
ad2dfb7180
Merge branch 'master' into nluehr/examplefixes
2017-02-15 14:29:27 -08:00
8e2f5efcf1
address review feedback, iteration 2
2017-02-14 13:08:37 -08:00
5f350c2c0f
address review comments
2017-02-13 16:05:35 -08:00
2789f5b0bb
Small language change to comments
...
I think you meant "Set epoch size to a larger number for lower training error" rather than "Set epoch size to a larger number of lower training error"
Thanks for the tutorials!
2017-02-12 10:34:48 -08:00
d04b5c5e56
remove old csv link
2017-02-10 15:22:03 -08:00
5d65ee6e9d
reflect beta11 api change
2017-02-10 13:52:38 -08:00
2f7a2f04c8
Cosmetic tweak in tutorials.
...
Make progress printer show correct epochs limit in most cases.
2017-02-10 13:19:39 -08:00
6f74b33696
CNTK_201B_CIFAR-10_ImageHandsOn.ipynb: fix PIL.Image import
2017-02-10 15:57:51 +01:00
Mark Hillebrand
Ivan Rodriguez
Project Philly
Alexey Reznichenko
Amit Agarwal
Emad Barsoum
Willi Richert
REDMOND\sayanpa
Frank Seide
YanZhangADS
William Darling
Michael Lanzetta
Frank Ibem
NORTHAMERICA\vistepan
Chris Basoglu
Guenther Schmuelling
Nikos Karampatziakis
jerogers-msft
Nathan Luehr