CNNSupervisedTrainer.ftl 15.9 KB
Newer Older
Nicola Gatto's avatar
Nicola Gatto committed
1
2
3
4
5
6
import mxnet as mx
import logging
import numpy as np
import time
import os
import shutil
7
import pickle
Sebastian N.'s avatar
Sebastian N. committed
8
9
import math
import sys
Nicola Gatto's avatar
Nicola Gatto committed
10
11
from mxnet import gluon, autograd, nd

Eyüp Harputlu's avatar
Eyüp Harputlu committed
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
class CrossEntropyLoss(gluon.loss.Loss):
    def __init__(self, axis=-1, sparse_label=True, weight=None, batch_axis=0, **kwargs):
        super(CrossEntropyLoss, self).__init__(weight, batch_axis, **kwargs)
        self._axis = axis
        self._sparse_label = sparse_label

    def hybrid_forward(self, F, pred, label, sample_weight=None):
        pred = F.log(pred)
        if self._sparse_label:
            loss = -F.pick(pred, label, axis=self._axis, keepdims=True)
        else:
            label = gluon.loss._reshape_like(F, label, pred)
            loss = -F.sum(pred * label, axis=self._axis, keepdims=True)
        loss = gluon.loss._apply_weighting(F, loss, self._weight, sample_weight)
        return F.mean(loss, axis=self._batch_axis, exclude=True)

Eyüp Harputlu's avatar
Eyüp Harputlu committed
28
29
30
31
32
33
34
35
36
class LogCoshLoss(gluon.loss.Loss):
    def __init__(self, weight=None, batch_axis=0, **kwargs):
        super(LogCoshLoss, self).__init__(weight, batch_axis, **kwargs)

    def hybrid_forward(self, F, pred, label, sample_weight=None):
        loss = F.log(F.cosh(pred - label))
        loss = gluon.loss._apply_weighting(F, loss, self._weight, sample_weight)
        return F.mean(loss, axis=self._batch_axis, exclude=True)

37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
class SoftmaxCrossEntropyLossIgnoreIndices(gluon.loss.Loss):
    def __init__(self, axis=-1, ignore_indices=[], sparse_label=True, from_logits=False, weight=None, batch_axis=0, **kwargs):
        super(SoftmaxCrossEntropyLossIgnoreIndices, self).__init__(weight, batch_axis, **kwargs)
        self._axis = axis
        self._ignore_indices = ignore_indices
        self._sparse_label = sparse_label
        self._from_logits = from_logits

    def hybrid_forward(self, F, pred, label, sample_weight=None):
        log_softmax = F.log_softmax
        pick = F.pick
        if not self._from_logits:
            pred = log_softmax(pred, self._axis)
        if self._sparse_label:
            loss = -pick(pred, label, axis=self._axis, keepdims=True)
        else:
            label = _reshape_like(F, label, pred)
            loss = -(pred * label).sum(axis=self._axis, keepdims=True)
        #loss = _apply_weighting(F, loss, self._weight, sample_weight)
        # ignore some indices for loss, e.g. <pad> tokens in NLP applications
        for i in self._ignore_indices:
            loss = loss * mx.nd.logical_not(mx.nd.equal(mx.nd.argmax(pred, axis=1), mx.nd.ones_like(mx.nd.argmax(pred, axis=1))*i))
        return loss.mean(axis=self._batch_axis, exclude=True)

Sebastian N.'s avatar
Sebastian N. committed
61
62
63
64
65
66
67
@mx.metric.register
class BLEU(mx.metric.EvalMetric):
    N = 4

    def __init__(self, exclude=None, name='bleu', output_names=None, label_names=None):
        super(BLEU, self).__init__(name=name, output_names=output_names, label_names=label_names)

Sebastian N.'s avatar
Sebastian N. committed
68
        self._exclude = exclude or []
Sebastian N.'s avatar
Sebastian N. committed
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169

        self._match_counts = [0 for _ in range(self.N)]
        self._counts = [0 for _ in range(self.N)]

        self._size_ref = 0
        self._size_hyp = 0

    def update(self, labels, preds):
        labels, preds = mx.metric.check_label_shapes(labels, preds, True)

        new_labels = self._convert(labels)
        new_preds = self._convert(preds)

        for label, pred in zip(new_labels, new_preds):
            reference = [word for word in label if word not in self._exclude]
            hypothesis = [word for word in pred if word not in self._exclude]

            self._size_ref += len(reference)
            self._size_hyp += len(hypothesis)

            for n in range(self.N):
                reference_ngrams = self._get_ngrams(reference, n + 1)
                hypothesis_ngrams = self._get_ngrams(hypothesis, n + 1)

                match_count = 0

                for ngram in hypothesis_ngrams:
                    if ngram in reference_ngrams:
                        reference_ngrams.remove(ngram)

                        match_count += 1

                self._match_counts[n] += match_count
                self._counts[n] += len(hypothesis_ngrams)

    def get(self):
        precisions = [sys.float_info.min for n in range(self.N)]

        i = 1

        for n in range(self.N):
            match_counts = self._match_counts[n]
            counts = self._counts[n]

            if counts != 0:
                if match_counts == 0:
                    i *= 2
                    match_counts = 1 / i

                precisions[n] = match_counts / counts

        bleu = self._get_brevity_penalty() * math.exp(sum(map(math.log, precisions)) / self.N)

        return (self.name, bleu)

    def calculate(self):
        precisions = [sys.float_info.min for n in range(self.N)]

        i = 1

        for n in range(self.N):
            match_counts = self._match_counts[n]
            counts = self._counts[n]

            if counts != 0:
                if match_counts == 0:
                    i *= 2
                    match_counts = 1 / i

                precisions[n] = match_counts / counts

        return self._get_brevity_penalty() * math.exp(sum(map(math.log, precisions)) / self.N)

    def _get_brevity_penalty(self):
        if self._size_hyp >= self._size_ref:
            return 1
        else:
            return math.exp(1 - (self._size_ref / self._size_hyp))

    @staticmethod
    def _get_ngrams(sentence, n):
        ngrams = []

        if len(sentence) >= n:
            for i in range(len(sentence) - n + 1):
                ngrams.append(sentence[i:i+n])

        return ngrams

    @staticmethod
    def _convert(nd_list):
        if len(nd_list) == 0:
            return []

        new_list = [[] for _ in range(nd_list[0].shape[0])]

        for element in nd_list:
            for i in range(element.shape[0]):
                new_list[i].append(element[i].asscalar())

        return new_list
170

171
172


173
class ${tc.fileNameWithoutEnding}:
Christian Fuß's avatar
Christian Fuß committed
174
    def applyBeamSearch(input, length, width, maxLength, currProb, netIndex, bestOutput):
175
        bestProb = 0.0
Christian Fuß's avatar
Christian Fuß committed
176
177
        while length < maxLength:
            length += 1
178
179
180
181
182
183
184
185
            batchIndex = 0
            for batchEntry in input:
                top_k_indices = mx.nd.topk(batchEntry, axis=0, k=width)
                top_k_values = mx.nd.topk(batchEntry, ret_typ='value', axis=0, k=width)
                for index in range(top_k_indices.size):

                    #print mx.nd.array(top_k_indices[index])
                    #print top_k_values[index]
Christian Fuß's avatar
Christian Fuß committed
186
                    if length == 1:
187
                        #print mx.nd.array(top_k_indices[index])
Christian Fuß's avatar
Christian Fuß committed
188
                        result = applyBeamSearch(self._networks[netIndex](mx.nd.array(top_k_indices[index])), length, width, maxLength,
189
190
                            currProb * top_k_values[index], netIndex, self._networks[netIndex](mx.nd.array(top_k_indices[index])))
                    else:
Christian Fuß's avatar
Christian Fuß committed
191
                        result = applyBeamSearch(self._networks[netIndex](mx.nd.array(top_k_indices[index])), length, width, maxLength,
192
193
                            currProb * top_k_values[index], netIndex, bestOutput)

Christian Fuß's avatar
Christian Fuß committed
194
                    if length == maxLength:
195
196
197
198
199
200
201
202
203
204
205
206
                        #print currProb
                        if currProb > bestProb:
                            bestProb = currProb
                            bestOutput[batchIndex] = result[batchIndex]
                            #print "new bestOutput: ", bestOutput

                batchIndex += 1
        #print bestOutput
        #print bestProb
        return bestOutput


207
    def __init__(self, data_loader, net_constructor):
Nicola Gatto's avatar
Nicola Gatto committed
208
209
        self._data_loader = data_loader
        self._net_creator = net_constructor
210
        self._networks = {}
Nicola Gatto's avatar
Nicola Gatto committed
211
212
213
214

    def train(self, batch_size=64,
              num_epoch=10,
              eval_metric='acc',
Sebastian N.'s avatar
Sebastian N. committed
215
              eval_metric_params={},
Eyüp Harputlu's avatar
Eyüp Harputlu committed
216
217
              loss ='softmax_cross_entropy',
              loss_params={},
Nicola Gatto's avatar
Nicola Gatto committed
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
              optimizer='adam',
              optimizer_params=(('learning_rate', 0.001),),
              load_checkpoint=True,
              context='gpu',
              checkpoint_period=5,
              normalize=True):
        if context == 'gpu':
            mx_context = mx.gpu()
        elif context == 'cpu':
            mx_context = mx.cpu()
        else:
            logging.error("Context argument is '" + context + "'. Only 'cpu' and 'gpu are valid arguments'.")

        if 'weight_decay' in optimizer_params:
            optimizer_params['wd'] = optimizer_params['weight_decay']
            del optimizer_params['weight_decay']
        if 'learning_rate_decay' in optimizer_params:
            min_learning_rate = 1e-08
            if 'learning_rate_minimum' in optimizer_params:
                min_learning_rate = optimizer_params['learning_rate_minimum']
                del optimizer_params['learning_rate_minimum']
            optimizer_params['lr_scheduler'] = mx.lr_scheduler.FactorScheduler(
                                                   optimizer_params['step_size'],
                                                   factor=optimizer_params['learning_rate_decay'],
                                                   stop_factor_lr=min_learning_rate)
            del optimizer_params['step_size']
            del optimizer_params['learning_rate_decay']


        train_iter, test_iter, data_mean, data_std = self._data_loader.load_data(batch_size)
248
249
250
251
252

        if normalize:
            self._net_creator.construct(context=mx_context, data_mean=data_mean, data_std=data_std)
        else:
            self._net_creator.construct(context=mx_context)
Nicola Gatto's avatar
Nicola Gatto committed
253
254
255
256
257
258
259
260

        begin_epoch = 0
        if load_checkpoint:
            begin_epoch = self._net_creator.load(mx_context)
        else:
            if os.path.isdir(self._net_creator._model_dir_):
                shutil.rmtree(self._net_creator._model_dir_)

261
        self._networks = self._net_creator.networks
Nicola Gatto's avatar
Nicola Gatto committed
262
263
264
265
266
267
268

        try:
            os.makedirs(self._net_creator._model_dir_)
        except OSError:
            if not os.path.isdir(self._net_creator._model_dir_):
                raise

Sebastian N.'s avatar
Sebastian N. committed
269
        trainers = [mx.gluon.Trainer(network.collect_params(), optimizer, optimizer_params) if len(net.collect_params().values()) != 0 for network in self._networks.values()]
Nicola Gatto's avatar
Nicola Gatto committed
270

Eyüp Harputlu's avatar
Eyüp Harputlu committed
271
272
273
274
275
276
        margin = loss_params['margin'] if 'margin' in loss_params else 1.0
        sparseLabel = loss_params['sparse_label'] if 'sparse_label' in loss_params else True
        if loss == 'softmax_cross_entropy':
            fromLogits = loss_params['from_logits'] if 'from_logits' in loss_params else False
            loss_function = mx.gluon.loss.SoftmaxCrossEntropyLoss(from_logits=fromLogits, sparse_label=sparseLabel)
        elif loss == 'sigmoid_binary_cross_entropy':
Nicola Gatto's avatar
Nicola Gatto committed
277
            loss_function = mx.gluon.loss.SigmoidBinaryCrossEntropyLoss()
Eyüp Harputlu's avatar
Eyüp Harputlu committed
278
279
280
        elif loss == 'cross_entropy':
            loss_function = CrossEntropyLoss(sparse_label=sparseLabel)
        elif loss == 'l2':
Nicola Gatto's avatar
Nicola Gatto committed
281
            loss_function = mx.gluon.loss.L2Loss()
Eyüp Harputlu's avatar
Eyüp Harputlu committed
282
        elif loss == 'l1':
Nicola Gatto's avatar
Nicola Gatto committed
283
            loss_function = mx.gluon.loss.L2Loss()
Eyüp Harputlu's avatar
Eyüp Harputlu committed
284
285
286
287
288
289
290
291
292
293
294
295
296
        elif loss == 'huber':
            rho = loss_params['rho'] if 'rho' in loss_params else 1
            loss_function = mx.gluon.loss.HuberLoss(rho=rho)
        elif loss == 'hinge':
            loss_function = mx.gluon.loss.HingeLoss(margin=margin)
        elif loss == 'squared_hinge':
            loss_function = mx.gluon.loss.SquaredHingeLoss(margin=margin)
        elif loss == 'logistic':
            labelFormat = loss_params['label_format'] if 'label_format' in loss_params else 'signed'
            loss_function = mx.gluon.loss.LogisticLoss(label_format=labelFormat)
        elif loss == 'kullback_leibler':
            fromLogits = loss_params['from_logits'] if 'from_logits' in loss_params else True
            loss_function = mx.gluon.loss.KLDivLoss(from_logits=fromLogits)
Eyüp Harputlu's avatar
Eyüp Harputlu committed
297
298
        elif loss == 'log_cosh':
            loss_function = LogCoshLoss()
Eyüp Harputlu's avatar
Eyüp Harputlu committed
299
300
        else:
            logging.error("Invalid loss parameter.")
Nicola Gatto's avatar
Nicola Gatto committed
301
302
303
304
305
306
307

        speed_period = 50
        tic = None

        for epoch in range(begin_epoch, begin_epoch + num_epoch):
            train_iter.reset()
            for batch_i, batch in enumerate(train_iter):
308
                <#list tc.architectureInputs as input_name>
309
                ${input_name} = batch.data[0].as_in_context(mx_context)
310
311
                </#list>
                <#list tc.architectureOutputs as output_name>
312
                ${output_name}label = batch.label[${output_name?index}].as_in_context(mx_context)
313
314
                </#list>

315
316
                outputs=[]

Nicola Gatto's avatar
Nicola Gatto committed
317
                with autograd.record():
318
<#include "pythonExecuteWithLoss.ftl">
319

320
321
322
                    loss = 0
                    for element in lossList:
                        loss = loss + element
323

Nicola Gatto's avatar
Nicola Gatto committed
324
                loss.backward()
325
326
327

                for trainer in trainers:
                    trainer.step(batch_size)
Nicola Gatto's avatar
Nicola Gatto committed
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344

                if tic is None:
                    tic = time.time()
                else:
                    if batch_i % speed_period == 0:
                        try:
                            speed = speed_period * batch_size / (time.time() - tic)
                        except ZeroDivisionError:
                            speed = float("inf")

                        logging.info("Epoch[%d] Batch[%d] Speed: %.2f samples/sec" % (epoch, batch_i, speed))

                        tic = time.time()

            tic = None

            train_iter.reset()
Sebastian N.'s avatar
Sebastian N. committed
345
            metric = mx.metric.create(eval_metric, **eval_metric_params)
Nicola Gatto's avatar
Nicola Gatto committed
346
            for batch_i, batch in enumerate(train_iter):
347
                <#list tc.architectureInputs as input_name>
348
                ${input_name} = batch.data[0].as_in_context(mx_context)
349
350
                </#list>

351
                labels = [
352
353
354
355
<#list tc.architectureOutputs as output_name>
                    batch.label[${output_name?index}].as_in_context(mx_context)<#sep>,
</#list>

356
357
                ]

358
                outputs=[]
359

360
361
                if True: <#-- Fix indentation -->
<#include "pythonExecute.ftl">
362

363
                predictions = []
364
                for output_name in outputs:
365
                    if mx.nd.shape_array(mx.nd.squeeze(output_name)).size > 1:
366
367
368
369
                        predictions.append(mx.nd.argmax(output_name, axis=1))
                    #ArgMax already applied
                    else:
                        predictions.append(output_name)
370

371
                metric.update(preds=predictions, labels=labels)
Nicola Gatto's avatar
Nicola Gatto committed
372
373
374
            train_metric_score = metric.get()[1]

            test_iter.reset()
Sebastian N.'s avatar
Sebastian N. committed
375
            metric = mx.metric.create(eval_metric, **eval_metric_params)
Nicola Gatto's avatar
Nicola Gatto committed
376
            for batch_i, batch in enumerate(test_iter):
377
                <#list tc.architectureInputs as input_name>
378
                ${input_name} = batch.data[0].as_in_context(mx_context)
379
380
                </#list>

381
                labels = [
382
383
384
385
<#list tc.architectureOutputs as output_name>
                    batch.label[${output_name?index}].as_in_context(mx_context)<#sep>,
</#list>

386
387
                ]

388
389
                outputs=[]

390
391
                if True: <#-- Fix indentation -->
<#include "pythonExecute.ftl">
392

393
                predictions = []
394
                for output_name in outputs:
395
                    if mx.nd.shape_array(mx.nd.squeeze(output_name)).size > 1:
396
397
398
399
                        predictions.append(mx.nd.argmax(output_name, axis=1))
                    #ArgMax already applied
                    else:
                        predictions.append(output_name)
400
401

                metric.update(preds=predictions, labels=labels)
Nicola Gatto's avatar
Nicola Gatto committed
402
403
404
405
            test_metric_score = metric.get()[1]

            logging.info("Epoch[%d] Train: %f, Test: %f" % (epoch, train_metric_score, test_metric_score))

406

Nicola Gatto's avatar
Nicola Gatto committed
407
            if (epoch - begin_epoch) % checkpoint_period == 0:
408
409
                for i, network in self._networks.items():
                    network.save_parameters(self.parameter_path(i) + '-' + str(epoch).zfill(4) + '.params')
Nicola Gatto's avatar
Nicola Gatto committed
410

411
412
413
        for i, network in self._networks.items():
            network.save_parameters(self.parameter_path(i) + '-' + str(num_epoch + begin_epoch).zfill(4) + '.params')
            network.export(self.parameter_path(i) + '_newest', epoch=0)
Nicola Gatto's avatar
Nicola Gatto committed
414

415
416
    def parameter_path(self, index):
        return self._net_creator._model_dir_ + self._net_creator._model_prefix_ + '_' + str(index)