aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorThomas Mesnard <thomas.mesnard@ens.fr>2015-12-20 10:58:07 +0100
committerThomas Mesnard <thomas.mesnard@ens.fr>2015-12-20 10:58:07 +0100
commite4d1c9c632158d7f40d26567d10c558c8815b84d (patch)
tree0e823d931ea18cbd69fbad05c7cd17baeba0c528
parent2142359b779519dab8b31a7b04c17c8539942a77 (diff)
downloadpgm-ctc-e4d1c9c632158d7f40d26567d10c558c8815b84d.tar.gz
pgm-ctc-e4d1c9c632158d7f40d26567d10c558c8815b84d.zip
Add some random code
-rw-r--r--.gitignore3
-rw-r--r--README.md2
-rw-r--r--mohammad/LICENSE201
-rw-r--r--mohammad/README.md33
-rw-r--r--mohammad/ctc_cost.py206
-rw-r--r--mohammad/ctc_test_data.pklbin0 -> 2080084 bytes
-rw-r--r--mohammad/test_ctc.py135
7 files changed, 580 insertions, 0 deletions
diff --git a/.gitignore b/.gitignore
new file mode 100644
index 0000000..5942bbc
--- /dev/null
+++ b/.gitignore
@@ -0,0 +1,3 @@
+__pycache__/
+*.pyc
+*.swp
diff --git a/README.md b/README.md
index 7dd39a7..c4c9da8 100644
--- a/README.md
+++ b/README.md
@@ -1,2 +1,4 @@
# pgm
Projet PGM
+
+Thomas Mesnard, Alex Auvolat
diff --git a/mohammad/LICENSE b/mohammad/LICENSE
new file mode 100644
index 0000000..5c304d1
--- /dev/null
+++ b/mohammad/LICENSE
@@ -0,0 +1,201 @@
+Apache License
+ Version 2.0, January 2004
+ http://www.apache.org/licenses/
+
+ TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+ 1. Definitions.
+
+ "License" shall mean the terms and conditions for use, reproduction,
+ and distribution as defined by Sections 1 through 9 of this document.
+
+ "Licensor" shall mean the copyright owner or entity authorized by
+ the copyright owner that is granting the License.
+
+ "Legal Entity" shall mean the union of the acting entity and all
+ other entities that control, are controlled by, or are under common
+ control with that entity. For the purposes of this definition,
+ "control" means (i) the power, direct or indirect, to cause the
+ direction or management of such entity, whether by contract or
+ otherwise, or (ii) ownership of fifty percent (50%) or more of the
+ outstanding shares, or (iii) beneficial ownership of such entity.
+
+ "You" (or "Your") shall mean an individual or Legal Entity
+ exercising permissions granted by this License.
+
+ "Source" form shall mean the preferred form for making modifications,
+ including but not limited to software source code, documentation
+ source, and configuration files.
+
+ "Object" form shall mean any form resulting from mechanical
+ transformation or translation of a Source form, including but
+ not limited to compiled object code, generated documentation,
+ and conversions to other media types.
+
+ "Work" shall mean the work of authorship, whether in Source or
+ Object form, made available under the License, as indicated by a
+ copyright notice that is included in or attached to the work
+ (an example is provided in the Appendix below).
+
+ "Derivative Works" shall mean any work, whether in Source or Object
+ form, that is based on (or derived from) the Work and for which the
+ editorial revisions, annotations, elaborations, or other modifications
+ represent, as a whole, an original work of authorship. For the purposes
+ of this License, Derivative Works shall not include works that remain
+ separable from, or merely link (or bind by name) to the interfaces of,
+ the Work and Derivative Works thereof.
+
+ "Contribution" shall mean any work of authorship, including
+ the original version of the Work and any modifications or additions
+ to that Work or Derivative Works thereof, that is intentionally
+ submitted to Licensor for inclusion in the Work by the copyright owner
+ or by an individual or Legal Entity authorized to submit on behalf of
+ the copyright owner. For the purposes of this definition, "submitted"
+ means any form of electronic, verbal, or written communication sent
+ to the Licensor or its representatives, including but not limited to
+ communication on electronic mailing lists, source code control systems,
+ and issue tracking systems that are managed by, or on behalf of, the
+ Licensor for the purpose of discussing and improving the Work, but
+ excluding communication that is conspicuously marked or otherwise
+ designated in writing by the copyright owner as "Not a Contribution."
+
+ "Contributor" shall mean Licensor and any individual or Legal Entity
+ on behalf of whom a Contribution has been received by Licensor and
+ subsequently incorporated within the Work.
+
+ 2. Grant of Copyright License. Subject to the terms and conditions of
+ this License, each Contributor hereby grants to You a perpetual,
+ worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+ copyright license to reproduce, prepare Derivative Works of,
+ publicly display, publicly perform, sublicense, and distribute the
+ Work and such Derivative Works in Source or Object form.
+
+ 3. Grant of Patent License. Subject to the terms and conditions of
+ this License, each Contributor hereby grants to You a perpetual,
+ worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+ (except as stated in this section) patent license to make, have made,
+ use, offer to sell, sell, import, and otherwise transfer the Work,
+ where such license applies only to those patent claims licensable
+ by such Contributor that are necessarily infringed by their
+ Contribution(s) alone or by combination of their Contribution(s)
+ with the Work to which such Contribution(s) was submitted. If You
+ institute patent litigation against any entity (including a
+ cross-claim or counterclaim in a lawsuit) alleging that the Work
+ or a Contribution incorporated within the Work constitutes direct
+ or contributory patent infringement, then any patent licenses
+ granted to You under this License for that Work shall terminate
+ as of the date such litigation is filed.
+
+ 4. Redistribution. You may reproduce and distribute copies of the
+ Work or Derivative Works thereof in any medium, with or without
+ modifications, and in Source or Object form, provided that You
+ meet the following conditions:
+
+ (a) You must give any other recipients of the Work or
+ Derivative Works a copy of this License; and
+
+ (b) You must cause any modified files to carry prominent notices
+ stating that You changed the files; and
+
+ (c) You must retain, in the Source form of any Derivative Works
+ that You distribute, all copyright, patent, trademark, and
+ attribution notices from the Source form of the Work,
+ excluding those notices that do not pertain to any part of
+ the Derivative Works; and
+
+ (d) If the Work includes a "NOTICE" text file as part of its
+ distribution, then any Derivative Works that You distribute must
+ include a readable copy of the attribution notices contained
+ within such NOTICE file, excluding those notices that do not
+ pertain to any part of the Derivative Works, in at least one
+ of the following places: within a NOTICE text file distributed
+ as part of the Derivative Works; within the Source form or
+ documentation, if provided along with the Derivative Works; or,
+ within a display generated by the Derivative Works, if and
+ wherever such third-party notices normally appear. The contents
+ of the NOTICE file are for informational purposes only and
+ do not modify the License. You may add Your own attribution
+ notices within Derivative Works that You distribute, alongside
+ or as an addendum to the NOTICE text from the Work, provided
+ that such additional attribution notices cannot be construed
+ as modifying the License.
+
+ You may add Your own copyright statement to Your modifications and
+ may provide additional or different license terms and conditions
+ for use, reproduction, or distribution of Your modifications, or
+ for any such Derivative Works as a whole, provided Your use,
+ reproduction, and distribution of the Work otherwise complies with
+ the conditions stated in this License.
+
+ 5. Submission of Contributions. Unless You explicitly state otherwise,
+ any Contribution intentionally submitted for inclusion in the Work
+ by You to the Licensor shall be under the terms and conditions of
+ this License, without any additional terms or conditions.
+ Notwithstanding the above, nothing herein shall supersede or modify
+ the terms of any separate license agreement you may have executed
+ with Licensor regarding such Contributions.
+
+ 6. Trademarks. This License does not grant permission to use the trade
+ names, trademarks, service marks, or product names of the Licensor,
+ except as required for reasonable and customary use in describing the
+ origin of the Work and reproducing the content of the NOTICE file.
+
+ 7. Disclaimer of Warranty. Unless required by applicable law or
+ agreed to in writing, Licensor provides the Work (and each
+ Contributor provides its Contributions) on an "AS IS" BASIS,
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+ implied, including, without limitation, any warranties or conditions
+ of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+ PARTICULAR PURPOSE. You are solely responsible for determining the
+ appropriateness of using or redistributing the Work and assume any
+ risks associated with Your exercise of permissions under this License.
+
+ 8. Limitation of Liability. In no event and under no legal theory,
+ whether in tort (including negligence), contract, or otherwise,
+ unless required by applicable law (such as deliberate and grossly
+ negligent acts) or agreed to in writing, shall any Contributor be
+ liable to You for damages, including any direct, indirect, special,
+ incidental, or consequential damages of any character arising as a
+ result of this License or out of the use or inability to use the
+ Work (including but not limited to damages for loss of goodwill,
+ work stoppage, computer failure or malfunction, or any and all
+ other commercial damages or losses), even if such Contributor
+ has been advised of the possibility of such damages.
+
+ 9. Accepting Warranty or Additional Liability. While redistributing
+ the Work or Derivative Works thereof, You may choose to offer,
+ and charge a fee for, acceptance of support, warranty, indemnity,
+ or other liability obligations and/or rights consistent with this
+ License. However, in accepting such obligations, You may act only
+ on Your own behalf and on Your sole responsibility, not on behalf
+ of any other Contributor, and only if You agree to indemnify,
+ defend, and hold each Contributor harmless for any liability
+ incurred by, or claims asserted against, such Contributor by reason
+ of your accepting any such warranty or additional liability.
+
+ END OF TERMS AND CONDITIONS
+
+ APPENDIX: How to apply the Apache License to your work.
+
+ To apply the Apache License to your work, attach the following
+ boilerplate notice, with the fields enclosed by brackets "{}"
+ replaced with your own identifying information. (Don't include
+ the brackets!) The text should be enclosed in the appropriate
+ comment syntax for the file format. We also recommend that a
+ file or class name and description of purpose be included on the
+ same "printed page" as the copyright notice for easier
+ identification within third-party archives.
+
+ Copyright {yyyy} {name of copyright owner}
+
+ Licensed under the Apache License, Version 2.0 (the "License");
+ you may not use this file except in compliance with the License.
+ You may obtain a copy of the License at
+
+ http://www.apache.org/licenses/LICENSE-2.0
+
+ Unless required by applicable law or agreed to in writing, software
+ distributed under the License is distributed on an "AS IS" BASIS,
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ See the License for the specific language governing permissions and
+ limitations under the License.
diff --git a/mohammad/README.md b/mohammad/README.md
new file mode 100644
index 0000000..8c553c3
--- /dev/null
+++ b/mohammad/README.md
@@ -0,0 +1,33 @@
+CTC-Connectionist Temporal Classification
+=========================================
+
+-CTC cost is implemented in pure [Theano](https://github.com/Theano/Theano).
+
+-Supports mini-batch.
+
+-Supports both normal- and log-scale.
+
+-"apple" problem (two same consecutive labels) is solved.
+
+-Test file is implemented using [Blocks](https://github.com/bartvm/blocks).
+
+
+
+
+Reference
+=========
+Graves, Alex, et al. "Connectionist temporal classification: labelling unsegmented sequence data with recurrent neural networks." Proceedings of the 23rd international conference on Machine learning. ACM, 2006.
+
+
+Credits
+=======
+Theano implementation of CTC by [Shawn Tan](https://github.com/shawntan/rnn-experiment/)
+
+Theano implementation of CTC by [Rakesh Var](https://github.com/rakeshvar/rnn_ctc)
+
+
+Special thanks to
+=================
+[Kyle Kastner](https://github.com/kastnerkyle)
+
+Pascal Lambdin
diff --git a/mohammad/ctc_cost.py b/mohammad/ctc_cost.py
new file mode 100644
index 0000000..979ed93
--- /dev/null
+++ b/mohammad/ctc_cost.py
@@ -0,0 +1,206 @@
+"""
+CTC-Connectionist Temporal Classification
+
+Code provided by Mohammad Pezeshki - May. 2015 -
+Montreal Institute for Learning Algorithms
+
+Referece: Graves, Alex, et al. "Connectionist temporal classification:
+labelling unsegmented sequence data with recurrent neural networks."
+Proceedings of the 23rd international conference on Machine learning.
+ACM, 2006.
+
+Credits: Shawn Tan, Rakesh Var
+
+This code is distributed without any warranty, express or implied.
+"""
+
+import theano
+from theano import tensor
+
+floatX = theano.config.floatX
+
+
+# T: INPUT_SEQUENCE_LENGTH
+# B: BATCH_SIZE
+# L: OUTPUT_SEQUENCE_LENGTH
+# C: NUM_CLASSES
+class CTC(object):
+ """Connectionist Temporal Classification
+ y_hat : T x B x C+1
+ y : L x B
+ y_hat_mask : T x B
+ y_mask : L x B
+ """
+ @staticmethod
+ def add_blanks(y, blank_symbol, y_mask=None):
+ """Add blanks to a matrix and updates mask
+
+ Input shape: L x B
+ Output shape: 2L+1 x B
+
+ """
+ # for y
+ y_extended = y.T.dimshuffle(0, 1, 'x')
+ blanks = tensor.zeros_like(y_extended) + blank_symbol
+ concat = tensor.concatenate([y_extended, blanks], axis=2)
+ res = concat.reshape((concat.shape[0],
+ concat.shape[1] * concat.shape[2])).T
+ begining_blanks = tensor.zeros((1, res.shape[1])) + blank_symbol
+ blanked_y = tensor.concatenate([begining_blanks, res], axis=0)
+ # for y_mask
+ if y_mask is not None:
+ y_mask_extended = y_mask.T.dimshuffle(0, 1, 'x')
+ concat = tensor.concatenate([y_mask_extended,
+ y_mask_extended], axis=2)
+ res = concat.reshape((concat.shape[0],
+ concat.shape[1] * concat.shape[2])).T
+ begining_blanks = tensor.ones((1, res.shape[1]), dtype=floatX)
+ blanked_y_mask = tensor.concatenate([begining_blanks, res], axis=0)
+ else:
+ blanked_y_mask = None
+ return blanked_y, blanked_y_mask
+
+ @staticmethod
+ def class_batch_to_labeling_batch(y, y_hat, y_hat_mask=None):
+ y_hat = y_hat * y_hat_mask.dimshuffle(0, 'x', 1)
+ batch_size = y_hat.shape[2]
+ res = y_hat[:, y.astype('int32'), tensor.arange(batch_size)]
+ return res
+
+ @staticmethod
+ def recurrence_relation(y, y_mask, blank_symbol):
+ n_y = y.shape[0]
+ blanks = tensor.zeros((2, y.shape[1])) + blank_symbol
+ ybb = tensor.concatenate((y, blanks), axis=0).T
+ sec_diag = (tensor.neq(ybb[:, :-2], ybb[:, 2:]) *
+ tensor.eq(ybb[:, 1:-1], blank_symbol) *
+ y_mask.T)
+
+ # r1: LxL
+ # r2: LxL
+ # r3: LxLxB
+ r2 = tensor.eye(n_y, k=1)
+ r3 = (tensor.eye(n_y, k=2).dimshuffle(0, 1, 'x') *
+ sec_diag.dimshuffle(1, 'x', 0))
+
+ return r2, r3
+
+ @classmethod
+ def path_probabs(cls, y, y_hat, y_mask, y_hat_mask, blank_symbol):
+ pred_y = cls.class_batch_to_labeling_batch(y, y_hat, y_hat_mask)
+
+ r2, r3 = cls.recurrence_relation(y, y_mask, blank_symbol)
+
+ def step(p_curr, p_prev):
+ # instead of dot product, we * first
+ # and then sum oven one dimension.
+ # objective: T.dot((p_prev)BxL, LxLxB)
+ # solusion: Lx1xB * LxLxB --> LxLxB --> (sumover)xLxB
+ dotproduct = (p_prev + tensor.dot(p_prev, r2) +
+ (p_prev.dimshuffle(1, 'x', 0) * r3).sum(axis=0).T)
+ return p_curr.T * dotproduct * y_mask.T # B x L
+
+ probabilities, _ = theano.scan(
+ step,
+ sequences=[pred_y],
+ outputs_info=[tensor.eye(y.shape[0])[0] * tensor.ones(y.T.shape)])
+ return probabilities, probabilities.shape
+
+ @classmethod
+ def cost(cls, y, y_hat, y_mask, y_hat_mask, blank_symbol):
+ y_hat_mask_len = tensor.sum(y_hat_mask, axis=0, dtype='int32')
+ y_mask_len = tensor.sum(y_mask, axis=0, dtype='int32')
+ probabilities, sth = cls.path_probabs(y, y_hat,
+ y_mask, y_hat_mask,
+ blank_symbol)
+ batch_size = probabilities.shape[1]
+ labels_probab = (probabilities[y_hat_mask_len - 1,
+ tensor.arange(batch_size),
+ y_mask_len - 1] +
+ probabilities[y_hat_mask_len - 1,
+ tensor.arange(batch_size),
+ y_mask_len - 2])
+ avg_cost = tensor.mean(-tensor.log(labels_probab))
+ return avg_cost, sth
+
+ @staticmethod
+ def _epslog(x):
+ return tensor.cast(tensor.log(tensor.clip(x, 1E-12, 1E12)),
+ theano.config.floatX)
+
+ @staticmethod
+ def log_add(a, b):
+ max_ = tensor.maximum(a, b)
+ return (max_ + tensor.log1p(tensor.exp(a + b - 2 * max_)))
+
+ @staticmethod
+ def log_dot_matrix(x, z):
+ inf = 1E12
+ log_dot = tensor.dot(x, z)
+ zeros_to_minus_inf = (z.max(axis=0) - 1) * inf
+ return log_dot + zeros_to_minus_inf
+
+ @staticmethod
+ def log_dot_tensor(x, z):
+ inf = 1E12
+ log_dot = (x.dimshuffle(1, 'x', 0) * z).sum(axis=0).T
+ zeros_to_minus_inf = (z.max(axis=0) - 1) * inf
+ return log_dot + zeros_to_minus_inf.T
+
+ @classmethod
+ def log_path_probabs(cls, y, y_hat, y_mask, y_hat_mask, blank_symbol):
+ pred_y = cls.class_batch_to_labeling_batch(y, y_hat, y_hat_mask)
+ r2, r3 = cls.recurrence_relation(y, y_mask, blank_symbol)
+
+ def step(log_p_curr, log_p_prev):
+ p1 = log_p_prev
+ p2 = cls.log_dot_matrix(p1, r2)
+ p3 = cls.log_dot_tensor(p1, r3)
+ p123 = cls.log_add(p3, cls.log_add(p1, p2))
+
+ return (log_p_curr.T +
+ p123 +
+ cls._epslog(y_mask.T))
+
+ log_probabilities, _ = theano.scan(
+ step,
+ sequences=[cls._epslog(pred_y)],
+ outputs_info=[cls._epslog(tensor.eye(y.shape[0])[0] *
+ tensor.ones(y.T.shape))])
+ return log_probabilities
+
+ @classmethod
+ def log_cost(cls, y, y_hat, y_mask, y_hat_mask, blank_symbol):
+ y_hat_mask_len = tensor.sum(y_hat_mask, axis=0, dtype='int32')
+ y_mask_len = tensor.sum(y_mask, axis=0, dtype='int32')
+ log_probabs = cls.log_path_probabs(y, y_hat,
+ y_mask, y_hat_mask,
+ blank_symbol)
+ batch_size = log_probabs.shape[1]
+ labels_probab = cls.log_add(
+ log_probabs[y_hat_mask_len - 1,
+ tensor.arange(batch_size),
+ y_mask_len - 1],
+ log_probabs[y_hat_mask_len - 1,
+ tensor.arange(batch_size),
+ y_mask_len - 2])
+ avg_cost = tensor.mean(-labels_probab)
+ return avg_cost
+
+ @classmethod
+ def apply(cls, y, y_hat, y_mask, y_hat_mask, scale='log_scale'):
+ y_hat = y_hat.dimshuffle(0, 2, 1)
+ num_classes = y_hat.shape[1] - 1
+ blanked_y, blanked_y_mask = cls.add_blanks(
+ y=y,
+ blank_symbol=num_classes.astype(floatX),
+ y_mask=y_mask)
+ if scale == 'log_scale':
+ final_cost = cls.log_cost(blanked_y, y_hat,
+ blanked_y_mask, y_hat_mask,
+ num_classes)
+ else:
+ final_cost, sth = cls.cost(blanked_y, y_hat,
+ blanked_y_mask, y_hat_mask,
+ num_classes)
+ return final_cost
diff --git a/mohammad/ctc_test_data.pkl b/mohammad/ctc_test_data.pkl
new file mode 100644
index 0000000..2833c1b
--- /dev/null
+++ b/mohammad/ctc_test_data.pkl
Binary files differ
diff --git a/mohammad/test_ctc.py b/mohammad/test_ctc.py
new file mode 100644
index 0000000..a24d634
--- /dev/null
+++ b/mohammad/test_ctc.py
@@ -0,0 +1,135 @@
+import theano
+import numpy
+from theano import tensor
+from blocks.model import Model
+from blocks.bricks import Linear, Tanh
+from ctc_cost import CTC
+from blocks.initialization import IsotropicGaussian, Constant
+from fuel.datasets import IterableDataset
+from fuel.streams import DataStream
+from blocks.algorithms import (GradientDescent, Scale,
+ StepClipping, CompositeRule)
+from blocks.extensions.monitoring import TrainingDataMonitoring
+from blocks.main_loop import MainLoop
+from blocks.extensions import FinishAfter, Printing
+from blocks.bricks.recurrent import SimpleRecurrent
+from blocks.graph import ComputationGraph
+try:
+ import cPickle as pickle
+except:
+ import pickle
+
+floatX = theano.config.floatX
+
+
+@theano.compile.ops.as_op(itypes=[tensor.lvector],
+ otypes=[tensor.lvector])
+def print_pred(y_hat):
+ blank_symbol = 4
+ res = []
+ for i, s in enumerate(y_hat):
+ if (s != blank_symbol) and (i == 0 or s != y_hat[i - 1]):
+ res += [s]
+ return numpy.asarray(res)
+
+n_epochs = 200
+x_dim = 4
+h_dim = 9
+num_classes = 4
+
+with open("ctc_test_data.pkl", "rb") as pkl_file:
+ try:
+ data = pickle.load(pkl_file)
+ inputs = data['inputs']
+ labels = data['labels']
+ # from S x T x B x D to S x T x B
+ inputs_mask = numpy.max(data['mask_inputs'], axis=-1)
+ labels_mask = data['mask_labels']
+ except:
+ data = pickle.load(pkl_file, encoding='bytes')
+ inputs = data[b'inputs']
+ labels = data[b'labels']
+ # from S x T x B x D to S x T x B
+ inputs_mask = numpy.max(data[b'mask_inputs'], axis=-1)
+ labels_mask = data[b'mask_labels']
+
+
+
+print('Building model ...')
+# T x B x F
+x = tensor.tensor3('x', dtype=floatX)
+# T x B
+x_mask = tensor.matrix('x_mask', dtype=floatX)
+# L x B
+y = tensor.matrix('y', dtype=floatX)
+# L x B
+y_mask = tensor.matrix('y_mask', dtype=floatX)
+
+x_to_h = Linear(name='x_to_h',
+ input_dim=x_dim,
+ output_dim=h_dim)
+x_transform = x_to_h.apply(x)
+rnn = SimpleRecurrent(activation=Tanh(),
+ dim=h_dim, name="rnn")
+h = rnn.apply(x_transform)
+h_to_o = Linear(name='h_to_o',
+ input_dim=h_dim,
+ output_dim=num_classes + 1)
+h_transform = h_to_o.apply(h)
+# T x B x C+1
+y_hat = tensor.nnet.softmax(
+ h_transform.reshape((-1, num_classes + 1))
+).reshape((h.shape[0], h.shape[1], -1))
+y_hat.name = 'y_hat'
+
+y_hat_mask = x_mask
+cost = CTC().apply(y, y_hat, y_mask, y_hat_mask, 'normal_scale')
+cost.name = 'CTC'
+# Initialization
+for brick in (rnn, x_to_h, h_to_o):
+ brick.weights_init = IsotropicGaussian(0.01)
+ brick.biases_init = Constant(0)
+ brick.initialize()
+
+print('Bulding DataStream ...')
+dataset = IterableDataset({'x': inputs,
+ 'x_mask': inputs_mask,
+ 'y': labels,
+ 'y_mask': labels_mask})
+stream = DataStream(dataset)
+
+print('Bulding training process...')
+algorithm = GradientDescent(cost=cost,
+ parameters=ComputationGraph(cost).parameters,
+ step_rule=CompositeRule([StepClipping(10.0),
+ Scale(0.02)]))
+monitor_cost = TrainingDataMonitoring([cost],
+ prefix="train",
+ after_epoch=True)
+
+# sample number to monitor
+sample = 8
+
+y_hat_max_path = print_pred(tensor.argmax(y_hat[:, sample, :], axis=1))
+y_hat_max_path.name = 'Viterbi'
+monitor_output = TrainingDataMonitoring([y_hat_max_path],
+ prefix="y_hat",
+ every_n_epochs=1)
+
+length = tensor.sum(y_mask[:, sample]).astype('int32')
+tar = y[:length, sample].astype('int32')
+tar.name = '_Target_Seq'
+monitor_target = TrainingDataMonitoring([tar],
+ prefix="y",
+ every_n_epochs=1)
+
+model = Model(cost)
+main_loop = MainLoop(data_stream=stream, algorithm=algorithm,
+ extensions=[monitor_cost, monitor_output,
+ monitor_target,
+ FinishAfter(after_n_epochs=n_epochs),
+ Printing()],
+ model=model)
+
+print('Starting training ...')
+main_loop.run()