File size: 11,415 Bytes
cd6775d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
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
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
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
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
# Copyright (c) 2019-present, Facebook, Inc.
# All rights reserved.
#
# This source code is licensed under the license found in the
# LICENSE file in the root directory of this source tree.
#

from logging import getLogger
import os
import copy
import time
import json
from collections import OrderedDict

import numpy as np
import torch
from torch import nn
import torch.nn.functional as F

from scipy.stats import spearmanr, pearsonr
from sklearn.metrics import f1_score, matthews_corrcoef

from ..optim import get_optimizer
from ..utils import concat_batches, truncate, to_cuda
from ..data.dataset import Dataset, ParallelDataset
from ..data.loader import load_binarized, set_dico_parameters


N_CLASSES = {
    'MNLI-m': 3,
    'MNLI-mm': 3,
    'QQP': 2,
    'QNLI': 2,
    'SST-2': 2,
    'CoLA': 2,
    'MRPC': 2,
    'RTE': 2,
    'STS-B': 1,
    'WNLI': 2,
    'AX_MNLI-m': 3,
}


logger = getLogger()


class GLUE:

    def __init__(self, embedder, scores, params):
        """
        Initialize GLUE trainer / evaluator.
        Initial `embedder` should be on CPU to save memory.
        """
        self._embedder = embedder
        self.params = params
        self.scores = scores

    def get_iterator(self, splt):
        """
        Build data iterator.
        """
        return self.data[splt]['x'].get_iterator(
            shuffle=(splt == 'train'),
            return_indices=True,
            group_by_size=self.params.group_by_size
        )

    def run(self, task):
        """
        Run GLUE training / evaluation.
        """
        params = self.params

        # task parameters
        self.task = task
        params.out_features = N_CLASSES[task]
        self.is_classif = task != 'STS-B'

        # load data
        self.data = self.load_data(task)
        if not self.data['dico'] == self._embedder.dico:
            raise Exception(("Dictionary in evaluation data (%i words) seems different than the one " +
                             "in the pretrained model (%i words). Please verify you used the same dictionary, " +
                             "and the same values for max_vocab and min_count.") % (len(self.data['dico']), len(self._embedder.dico)))

        # embedder
        self.embedder = copy.deepcopy(self._embedder)
        self.embedder.cuda()

        # projection layer
        self.proj = nn.Sequential(*[
            nn.Dropout(params.dropout),
            nn.Linear(self.embedder.out_dim, params.out_features)
        ]).cuda()

        # optimizers
        self.optimizer_e = get_optimizer(list(self.embedder.get_parameters(params.finetune_layers)), params.optimizer_e)
        self.optimizer_p = get_optimizer(self.proj.parameters(), params.optimizer_p)

        # train and evaluate the model
        for epoch in range(params.n_epochs):

            # update epoch
            self.epoch = epoch

            # training
            logger.info("GLUE - %s - Training epoch %i ..." % (task, epoch))
            self.train()

            # evaluation
            logger.info("GLUE - %s - Evaluating epoch %i ..." % (task, epoch))
            with torch.no_grad():
                scores = self.eval('valid')
                self.scores.update(scores)
                self.eval('test')

    def train(self):
        """
        Finetune for one epoch on the training set.
        """
        params = self.params
        self.embedder.train()
        self.proj.train()

        # training variables
        losses = []
        ns = 0  # number of sentences
        nw = 0  # number of words
        t = time.time()

        iterator = self.get_iterator('train')
        lang_id = params.lang2id['en']

        while True:

            # batch
            try:
                batch = next(iterator)
            except StopIteration:
                break
            if self.n_sent == 1:
                (x, lengths), idx = batch
                x, lengths = truncate(x, lengths, params.max_len, params.eos_index)
            else:
                (sent1, len1), (sent2, len2), idx = batch
                sent1, len1 = truncate(sent1, len1, params.max_len, params.eos_index)
                sent2, len2 = truncate(sent2, len2, params.max_len, params.eos_index)
                x, lengths, _, _ = concat_batches(sent1, len1, lang_id, sent2, len2, lang_id, params.pad_index, params.eos_index, reset_positions=False)
            y = self.data['train']['y'][idx]
            bs = len(lengths)

            # cuda
            x, y, lengths = to_cuda(x, y, lengths)

            # loss
            output = self.proj(self.embedder.get_embeddings(x, lengths, positions=None, langs=None))
            if self.is_classif:
                loss = F.cross_entropy(output, y, weight=self.weights)
            else:
                loss = F.mse_loss(output.squeeze(1), y.float())

            # backward / optimization
            self.optimizer_e.zero_grad()
            self.optimizer_p.zero_grad()
            loss.backward()
            self.optimizer_e.step()
            self.optimizer_p.step()

            # update statistics
            ns += bs
            nw += lengths.sum().item()
            losses.append(loss.item())

            # log
            if ns != 0 and ns % (10 * bs) < bs:
                logger.info(
                    "GLUE - %s - Epoch %s - Train iter %7i - %.1f words/s - %s Loss: %.4f"
                    % (self.task, self.epoch, ns, nw / (time.time() - t), 'XE' if self.is_classif else 'MSE', sum(losses) / len(losses))
                )
                nw, t = 0, time.time()
                losses = []

            # epoch size
            if params.epoch_size != -1 and ns >= params.epoch_size:
                break

    def eval(self, splt):
        """
        Evaluate on XNLI validation and test sets, for all languages.
        """
        params = self.params
        self.embedder.eval()
        self.proj.eval()

        assert splt in ['valid', 'test']
        has_labels = 'y' in self.data[splt]

        scores = OrderedDict({'epoch': self.epoch})
        task = self.task.lower()

        idxs = []  # sentence indices
        prob = []  # probabilities
        pred = []  # predicted values
        gold = []  # real values

        lang_id = params.lang2id['en']

        for batch in self.get_iterator(splt):

            # batch
            if self.n_sent == 1:
                (x, lengths), idx = batch
                # x, lengths = truncate(x, lengths, params.max_len, params.eos_index)
            else:
                (sent1, len1), (sent2, len2), idx = batch
                # sent1, len1 = truncate(sent1, len1, params.max_len, params.eos_index)
                # sent2, len2 = truncate(sent2, len2, params.max_len, params.eos_index)
                x, lengths, _, _ = concat_batches(sent1, len1, lang_id, sent2, len2, lang_id, params.pad_index, params.eos_index, reset_positions=False)
            y = self.data[splt]['y'][idx] if has_labels else None

            # cuda
            x, y, lengths = to_cuda(x, y, lengths)

            # prediction
            output = self.proj(self.embedder.get_embeddings(x, lengths, positions=None, langs=None))
            p = output.data.max(1)[1] if self.is_classif else output.squeeze(1)
            idxs.append(idx)
            prob.append(output.cpu().numpy())
            pred.append(p.cpu().numpy())
            if has_labels:
                gold.append(y.cpu().numpy())

        # indices / predictions
        idxs = np.concatenate(idxs)
        prob = np.concatenate(prob)
        pred = np.concatenate(pred)
        assert len(idxs) == len(pred), (len(idxs), len(pred))
        assert idxs[-1] == len(idxs) - 1, (idxs[-1], len(idxs) - 1)

        # score the predictions if we have labels
        if has_labels:
            gold = np.concatenate(gold)
            prefix = f'{splt}_{task}'
            if self.is_classif:
                scores['%s_acc' % prefix] = 100. * (pred == gold).sum() / len(pred)
                scores['%s_f1' % prefix] = 100. * f1_score(gold, pred, average='binary' if params.out_features == 2 else 'micro')
                scores['%s_mc' % prefix] = 100. * matthews_corrcoef(gold, pred)
            else:
                scores['%s_prs' % prefix] = 100. * pearsonr(pred, gold)[0]
                scores['%s_spr' % prefix] = 100. * spearmanr(pred, gold)[0]
            logger.info("__log__:%s" % json.dumps(scores))

        # output predictions
        pred_path = os.path.join(params.dump_path, f'{splt}.pred.{self.epoch}')
        with open(pred_path, 'w') as f:
            for i, p in zip(idxs, prob):
                f.write('%i\t%s\n' % (i, ','.join([str(x) for x in p])))
        logger.info(f"Wrote {len(idxs)} {splt} predictions to {pred_path}")

        return scores

    def load_data(self, task):
        """
        Load pair regression/classification bi-sentence tasks
        """
        params = self.params
        data = {splt: {} for splt in ['train', 'valid', 'test']}
        dpath = os.path.join(params.data_path, 'eval', task)

        self.n_sent = 1 if task in ['SST-2', 'CoLA'] else 2

        for splt in ['train', 'valid', 'test']:

            # load data and dictionary
            data1 = load_binarized(os.path.join(dpath, '%s.s1.pth' % splt), params)
            data2 = load_binarized(os.path.join(dpath, '%s.s2.pth' % splt), params) if self.n_sent == 2 else None
            data['dico'] = data.get('dico', data1['dico'])

            # set dictionary parameters
            set_dico_parameters(params, data, data1['dico'])
            if self.n_sent == 2:
                set_dico_parameters(params, data, data2['dico'])

            # create dataset
            if self.n_sent == 1:
                data[splt]['x'] = Dataset(data1['sentences'], data1['positions'], params)
            else:
                data[splt]['x'] = ParallelDataset(
                    data1['sentences'], data1['positions'],
                    data2['sentences'], data2['positions'],
                    params
                )

            # load labels
            if splt != 'test' or task in ['MRPC']:
                # read labels from file
                with open(os.path.join(dpath, '%s.label' % splt), 'r') as f:
                    lines = [l.rstrip() for l in f]
                # STS-B task
                if task == 'STS-B':
                    assert all(0 <= float(x) <= 5 for x in lines)
                    y = [float(l) for l in lines]
                # QQP
                elif task == 'QQP':
                    UNK_LABEL = 0
                    lab2id = {x: i for i, x in enumerate(sorted(set(lines) - set([''])))}
                    y = [lab2id.get(x, UNK_LABEL) for x in lines]
                # other tasks
                else:
                    lab2id = {x: i for i, x in enumerate(sorted(set(lines)))}
                    y = [lab2id[x] for x in lines]
                data[splt]['y'] = torch.LongTensor(y)
                assert len(data[splt]['x']) == len(data[splt]['y'])

        # compute weights for weighted training
        if task != 'STS-B' and params.weighted_training:
            weights = torch.FloatTensor([
                1.0 / (data['train']['y'] == i).sum().item()
                for i in range(len(lab2id))
            ]).cuda()
            self.weights = weights / weights.sum()
        else:
            self.weights = None

        return data