Skip to content

Commit 75cbba6

Browse files
abergeronabergeron
committed
Python 3 + flake8 fixes.
1 parent 1f5c106 commit 75cbba6

File tree

4 files changed

+70
-77
lines changed

4 files changed

+70
-77
lines changed

‎code/DBN.py

Lines changed: 48 additions & 53 deletions
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,6 @@
11
"""
22
"""
3+
from __future__ import print_function, division
34
import os
45
import sys
56
import timeit
@@ -61,9 +62,12 @@ def __init__(self, numpy_rng, theano_rng=None, n_ins=784,
6162
theano_rng = MRG_RandomStreams(numpy_rng.randint(2 ** 30))
6263

6364
# allocate symbolic variables for the data
64-
self.x = T.matrix('x') # the data is presented as rasterized images
65-
self.y = T.ivector('y') # the labels are presented as 1D vector
66-
# of [int] labels
65+
66+
# the data is presented as rasterized images
67+
self.x = T.matrix('x')
68+
69+
# the labels are presented as 1D vector of [int] labels
70+
self.y = T.ivector('y')
6771
# end-snippet-1
6872
# The DBN is an MLP, for which all weights of intermediate
6973
# layers are shared with a different RBM. We will first
@@ -156,8 +160,6 @@ def pretraining_functions(self, train_set_x, batch_size, k):
156160
index = T.lscalar('index') # index to a minibatch
157161
learning_rate = T.scalar('lr') # learning rate to use
158162

159-
# number of batches
160-
n_batches = train_set_x.get_value(borrow=True).shape[0] / batch_size
161163
# begining of a batch, given `index`
162164
batch_begin = index * batch_size
163165
# ending of a batch given `index`
@@ -211,9 +213,9 @@ def build_finetune_functions(self, datasets, batch_size, learning_rate):
211213

212214
# compute number of minibatches for training, validation and testing
213215
n_valid_batches = valid_set_x.get_value(borrow=True).shape[0]
214-
n_valid_batches /= batch_size
216+
n_valid_batches //= batch_size
215217
n_test_batches = test_set_x.get_value(borrow=True).shape[0]
216-
n_test_batches /= batch_size
218+
n_test_batches //= batch_size
217219

218220
index = T.lscalar('index') # index to a [mini]batch
219221

@@ -307,11 +309,11 @@ def test_DBN(finetune_lr=0.1, pretraining_epochs=100,
307309
test_set_x, test_set_y = datasets[2]
308310

309311
# compute number of minibatches for training, validation and testing
310-
n_train_batches = train_set_x.get_value(borrow=True).shape[0] / batch_size
312+
n_train_batches = train_set_x.get_value(borrow=True).shape[0] // batch_size
311313

312314
# numpy random generator
313315
numpy_rng = numpy.random.RandomState(123)
314-
print '... building the model'
316+
print('... building the model')
315317
# construct the Deep Belief Network
316318
dbn = DBN(numpy_rng=numpy_rng, n_ins=28 * 28,
317319
hidden_layers_sizes=[1000, 1000, 1000],
@@ -321,14 +323,14 @@ def test_DBN(finetune_lr=0.1, pretraining_epochs=100,
321323
#########################
322324
# PRETRAINING THE MODEL #
323325
#########################
324-
print '... getting the pretraining functions'
326+
print('... getting the pretraining functions')
325327
pretraining_fns = dbn.pretraining_functions(train_set_x=train_set_x,
326328
batch_size=batch_size,
327329
k=k)
328330

329-
print '... pre-training the model'
331+
print('... pre-training the model')
330332
start_time = timeit.default_timer()
331-
## Pre-train layer-wise
333+
# Pre-train layer-wise
332334
for i in range(dbn.n_layers):
333335
# go through pretraining epochs
334336
for epoch in range(pretraining_epochs):
@@ -337,38 +339,40 @@ def test_DBN(finetune_lr=0.1, pretraining_epochs=100,
337339
for batch_index in range(n_train_batches):
338340
c.append(pretraining_fns[i](index=batch_index,
339341
lr=pretrain_lr))
340-
print 'Pre-training layer %i, epoch %d, cost ' % (i, epoch),
341-
print numpy.mean(c)
342+
print('Pre-training layer %i, epoch %d, cost ' % (i, epoch), end=' ')
343+
print(numpy.mean(c))
342344

343345
end_time = timeit.default_timer()
344346
# end-snippet-2
345-
print >> sys.stderr, ('The pretraining code for file ' +
346-
os.path.split(__file__)[1] +
347-
' ran for %.2fm' % ((end_time - start_time) / 60.))
347+
print('The pretraining code for file ' + os.path.split(__file__)[1] +
348+
' ran for %.2fm' % ((end_time - start_time) / 60.), file=sys.stderr)
348349
########################
349350
# FINETUNING THE MODEL #
350351
########################
351352

352353
# get the training, validation and testing function for the model
353-
print '... getting the finetuning functions'
354+
print('... getting the finetuning functions')
354355
train_fn, validate_model, test_model = dbn.build_finetune_functions(
355356
datasets=datasets,
356357
batch_size=batch_size,
357358
learning_rate=finetune_lr
358359
)
359360

360-
print '... finetuning the model'
361+
print('... finetuning the model')
361362
# early-stopping parameters
362-
patience = 4 * n_train_batches # look as this many examples regardless
363-
patience_increase = 2. # wait this much longer when a new best is
364-
# found
365-
improvement_threshold = 0.995 # a relative improvement of this much is
366-
# considered significant
363+
364+
# look as this many examples regardless
365+
patience = 4 * n_train_batches
366+
367+
# wait this much longer when a new best is found
368+
patience_increase = 2.
369+
370+
# a relative improvement of this much is considered significant
371+
improvement_threshold = 0.995
372+
373+
# go through this many minibatches before checking the network on
374+
# the validation set; in this case we check every epoch
367375
validation_frequency = min(n_train_batches, patience / 2)
368-
# go through this many
369-
# minibatches before checking the network
370-
# on the validation set; in this case we
371-
# check every epoch
372376

373377
best_validation_loss = numpy.inf
374378
test_score = 0.
@@ -381,31 +385,27 @@ def test_DBN(finetune_lr=0.1, pretraining_epochs=100,
381385
epoch = epoch + 1
382386
for minibatch_index in range(n_train_batches):
383387

384-
minibatch_avg_cost = train_fn(minibatch_index)
388+
train_fn(minibatch_index)
385389
iter = (epoch - 1) * n_train_batches + minibatch_index
386390

387391
if (iter + 1) % validation_frequency == 0:
388392

389393
validation_losses = validate_model()
390394
this_validation_loss = numpy.mean(validation_losses)
391-
print(
392-
'epoch %i, minibatch %i/%i, validation error %f %%'
393-
% (
394-
epoch,
395-
minibatch_index + 1,
396-
n_train_batches,
397-
this_validation_loss * 100.
395+
print('epoch %i, minibatch %i/%i, validation error %f %%' % (
396+
epoch,
397+
minibatch_index + 1,
398+
n_train_batches,
399+
this_validation_loss * 100.
398400
)
399401
)
400402

401403
# if we got the best validation score until now
402404
if this_validation_loss < best_validation_loss:
403405

404-
#improve patience if loss improvement is good enough
405-
if (
406-
this_validation_loss < best_validation_loss *
407-
improvement_threshold
408-
):
406+
# improve patience if loss improvement is good enough
407+
if (this_validation_loss < best_validation_loss *
408+
improvement_threshold):
409409
patience = max(patience, iter * patience_increase)
410410

411411
# save best validation score and iteration number
@@ -418,24 +418,19 @@ def test_DBN(finetune_lr=0.1, pretraining_epochs=100,
418418
print((' epoch %i, minibatch %i/%i, test error of '
419419
'best model %f %%') %
420420
(epoch, minibatch_index + 1, n_train_batches,
421-
test_score * 100.))
421+
test_score * 100.))
422422

423423
if patience <= iter:
424424
done_looping = True
425425
break
426426

427427
end_time = timeit.default_timer()
428-
print(
429-
(
430-
'Optimization complete with best validation score of %f %%, '
431-
'obtained at iteration %i, '
432-
'with test performance %f %%'
433-
) % (best_validation_loss * 100., best_iter + 1, test_score * 100.)
434-
)
435-
print >> sys.stderr, ('The fine tuning code for file ' +
436-
os.path.split(__file__)[1] +
437-
' ran for %.2fm' % ((end_time - start_time)
438-
/ 60.))
428+
print(('Optimization complete with best validation score of %f %%, '
429+
'obtained at iteration %i, '
430+
'with test performance %f %%'
431+
) % (best_validation_loss * 100., best_iter + 1, test_score * 100.))
432+
print('The fine tuning code for file ' + os.path.split(__file__)[1] +
433+
' ran for %.2fm' % ((end_time - start_time) / 60.), file=sys.stderr)
439434

440435

441436
if __name__ == '__main__':

‎code/imdb_preprocess.py

Lines changed: 5 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -8,7 +8,7 @@
88
99
3) Then run this script.
1010
"""
11-
11+
from __future__ import print_function
1212
dataset_path='/Tmp/bastienf/aclImdb/'
1313

1414
import numpy
@@ -27,12 +27,12 @@
2727

2828
def tokenize(sentences):
2929

30-
print 'Tokenizing..',
30+
print('Tokenizing..', end=' ')
3131
text = "\n".join(sentences)
3232
tokenizer = Popen(tokenizer_cmd, stdin=PIPE, stdout=PIPE)
3333
tok_text, _ = tokenizer.communicate(text)
3434
toks = tok_text.split('\n')[:-1]
35-
print 'Done'
35+
print('Done')
3636

3737
return toks
3838

@@ -52,7 +52,7 @@ def build_dict(path):
5252

5353
sentences = tokenize(sentences)
5454

55-
print 'Building dictionary..',
55+
print('Building dictionary..', end=' ')
5656
wordcount = dict()
5757
for ss in sentences:
5858
words = ss.strip().lower().split()
@@ -72,7 +72,7 @@ def build_dict(path):
7272
for idx, ss in enumerate(sorted_idx):
7373
worddict[keys[ss]] = idx+2 # leave 0 and 1 (UNK)
7474

75-
print numpy.sum(counts), ' total words ', len(keys), ' unique words'
75+
print(numpy.sum(counts), ' total words ', len(keys), ' unique words')
7676

7777
return worddict
7878

‎code/logistic_cg.py

Lines changed: 11 additions & 14 deletions
Original file line numberDiff line numberDiff line change
@@ -33,6 +33,7 @@
3333
3434
3535
"""
36+
from __future__ import print_function, division
3637
__docformat__ = 'restructedtext en'
3738

3839

@@ -165,17 +166,17 @@ def cg_optimization_mnist(n_epochs=50, mnist_pkl_gz='mnist.pkl.gz'):
165166

166167
batch_size = 600 # size of the minibatch
167168

168-
n_train_batches = train_set_x.get_value(borrow=True).shape[0] / batch_size
169-
n_valid_batches = valid_set_x.get_value(borrow=True).shape[0] / batch_size
170-
n_test_batches = test_set_x.get_value(borrow=True).shape[0] / batch_size
169+
n_train_batches = train_set_x.get_value(borrow=True).shape[0] // batch_size
170+
n_valid_batches = valid_set_x.get_value(borrow=True).shape[0] // batch_size
171+
n_test_batches = test_set_x.get_value(borrow=True).shape[0] // batch_size
171172

172173
n_in = 28 * 28 # number of input units
173174
n_out = 10 # number of output units
174175

175176
######################
176177
# BUILD ACTUAL MODEL #
177178
######################
178-
print '... building the model'
179+
print('... building the model')
179180

180181
# allocate symbolic variables for the data
181182
minibatch_offset = T.lscalar() # offset to the start of a [mini]batch
@@ -260,7 +261,7 @@ def callback(theta_value):
260261
validation_losses = [validate_model(i * batch_size)
261262
for i in range(n_valid_batches)]
262263
this_validation_loss = numpy.mean(validation_losses)
263-
print('validation error %f %%' % (this_validation_loss * 100.,))
264+
print(('validation error %f %%' % (this_validation_loss * 100.,)))
264265

265266
# check if it is better then best validation score got until now
266267
if this_validation_loss < validation_scores[0]:
@@ -288,17 +289,13 @@ def callback(theta_value):
288289
maxiter=n_epochs
289290
)
290291
end_time = timeit.default_timer()
291-
print(
292-
(
293-
'Optimization complete with best validation score of %f %%, with '
294-
'test performance %f %%'
295-
)
296-
% (validation_scores[0] * 100., validation_scores[1] * 100.)
292+
print(('Optimization complete with best validation score of %f %%, with '
293+
'test performance %f %%'
294+
) % (validation_scores[0] * 100., validation_scores[1] * 100.)
297295
)
298296

299-
print >> sys.stderr, ('The code for file ' +
300-
os.path.split(__file__)[1] +
301-
' ran for %.1fs' % ((end_time - start_time)))
297+
print('The code for file ' + os.path.split(__file__)[1] +
298+
' ran for %.1fs' % (end_time - start_time), file=sys.stderr)
302299

303300

304301
if __name__ == '__main__':

‎code/test.py

Lines changed: 6 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,4 @@
1+
from __future__ import absolute_import, print_function, division
12
import sys
23

34
import numpy
@@ -137,12 +138,12 @@ def speed():
137138
def time_test(m, l, idx, f, **kwargs):
138139
if not to_exec[idx]:
139140
return
140-
print algo[idx]
141+
print(algo[idx])
141142
ts = m.call_time
142143
try:
143144
f(**kwargs)
144-
except Exception, e:
145-
print >> sys.stderr, 'test', algo[idx], 'FAILED', e
145+
except Exception as e:
146+
print('test', algo[idx], 'FAILED', e, file=sys.stderr)
146147
l.append(numpy.nan)
147148
return
148149
te = m.call_time
@@ -265,7 +266,7 @@ def do_tests():
265266
print >> sys.stderr, 'gpu % expected/get', (
266267
expected_times_gpu / gpu_times)
267268

268-
print
269+
print()
269270
if do_float64 and do_float32:
270271
print >> sys.stderr, 'float64/float32', (
271272
float64_times / float32_times)
@@ -286,7 +287,7 @@ def compare(x, y):
286287
# time and the real time, we consider this an error.
287288
return sum((ratio < 0.95) + (ratio > 1.05))
288289

289-
print
290+
print()
290291
if do_float64:
291292
err = compare(expected_times_64, float64_times)
292293
print >> sys.stderr, 'speed_failure_float64=' + str(err)

0 commit comments

Comments
 (0)