def run()

in senteval/snli.py [0:0]

• 39 lines of code
• 9 McCabe index (conditional complexity)

    def run(self, params, batcher):
        self.X, self.y = {}, {}
        dico_label = {'entailment': 0,  'neutral': 1, 'contradiction': 2}
        for key in self.data:
            if key not in self.X:
                self.X[key] = []
            if key not in self.y:
                self.y[key] = []

            input1, input2, mylabels = self.data[key]
            enc_input = []
            n_labels = len(mylabels)
            for ii in range(0, n_labels, params.batch_size):
                batch1 = input1[ii:ii + params.batch_size]
                batch2 = input2[ii:ii + params.batch_size]

                if len(batch1) == len(batch2) and len(batch1) > 0:
                    enc1 = batcher(params, batch1)
                    enc2 = batcher(params, batch2)
                    enc_input.append(np.hstack((enc1, enc2, enc1 * enc2,
                                                np.abs(enc1 - enc2))))
                if (ii*params.batch_size) % (20000*params.batch_size) == 0:
                    logging.info("PROGRESS (encoding): %.2f%%" %
                                 (100 * ii / n_labels))
            self.X[key] = np.vstack(enc_input)
            self.y[key] = [dico_label[y] for y in mylabels]

        config = {'nclasses': 3, 'seed': self.seed,
                  'usepytorch': params.usepytorch,
                  'cudaEfficient': True,
                  'nhid': params.nhid, 'noreg': True}

        config_classifier = copy.deepcopy(params.classifier)
        config_classifier['max_epoch'] = 15
        config_classifier['epoch_size'] = 1
        config['classifier'] = config_classifier

        clf = SplitClassifier(self.X, self.y, config)
        devacc, testacc = clf.run()
        logging.debug('Dev acc : {0} Test acc : {1} for SNLI\n'
                      .format(devacc, testacc))
        return {'devacc': devacc, 'acc': testacc,
                'ndev': len(self.data['valid'][0]),
                'ntest': len(self.data['test'][0])}