test_avgloss, test_pred_res, _ = evaluate(model, test_data, loss_Function, word_to_ix, all_losses_test, 'test') return test_pred_res, all_losses, all_losses_dev if __name__ == '__main__': emotionlist = ['joy', 'anger', 'fear', 'sadness'] #emotionlist = ['joy', 'anger', 'fear', 'sadness'] #emotionlist = ['joy'] seedlist = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10] for emotion in emotionlist: trainfile = 'data/EI-reg-En-train/EI-reg-En-' + emotion + '-train.txt' devfile = 'data/2018-EI-reg-En-dev/2018-EI-reg-En-' + emotion + '-dev.txt' # testfile = 'data/2018-EI-reg-En-test/2018-EI-reg-En-'+emotion+'-test.txt' testfile = 'data/SemEval2018-Task1-AIT-Test-gold/EI-reg/2018-EI-reg-En-' + emotion + '-test-gold.txt' #train_data, dev_data, test_data, word_to_ix = dataLoaderRegresser.loadData(trainfile, devfile, testfile) train_data, dev_data, test_data, word_to_ix, char_to_ix = dataLoaderRegresser.loadDataChar(trainfile, devfile, testfile) for SEED in seedlist: torch.manual_seed(SEED) random.seed(SEED) print('EMOTION:', emotion, 'SEED:', SEED) train_dev_data = train_data + dev_data random.shuffle(train_dev_data) train_data = train_dev_data[:int(len(train_dev_data)*0.9)] dev_data = train_dev_data[int(len(train_dev_data)*0.9):] #test_data = test_data[:1106] print('-> len(test_data): ', len(test_data)) print('-> test_data example:', test_data[0]) print('-> test_data example: ', test_data[-1]) #EMBEDDING_DIM = 50
if __name__ == '__main__': emotionlist = ['joy', 'anger', 'fear', 'sadness'] #emotionlist = ['joy', 'anger', 'fear', 'sadness'] #emotionlist = ['joy'] seedlist = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10] for emotion in emotionlist: print('English data:') trainfile = 'data/EI-reg-En-train/EI-reg-En-' + emotion + '-train.txt' devfile = 'data/2018-EI-reg-En-dev/2018-EI-reg-En-' + emotion + '-dev.txt' # testfile = 'data/2018-EI-reg-En-test/2018-EI-reg-En-'+emotion+'-test.txt' testfile = 'data/SemEval2018-Task1-AIT-Test-gold/EI-reg/2018-EI-reg-En-' + emotion + '-test-gold.txt' #train_data, dev_data, test_data, word_to_ix = dataLoaderRegresser.loadData(trainfile, devfile, testfile) train_data, dev_data, test_data, word_to_ix, char_to_ix = dataLoaderRegresser.loadDataChar( trainfile, devfile, testfile) print('Arabic translated data:') trainfileAr = 'translated/Ar_' + emotion + '_train.txt' devfileAr = 'translated/Ar_' + emotion + '_dev.txt' testfileAr = 'translated/Ar_' + emotion + '_test.txt' train_dataAr, dev_dataAr, test_dataAr, word_to_ixAr, char_to_ixAr = dataLoaderRegresser.loadDataChar( trainfileAr, devfileAr, testfileAr) train_dev_dataAr = train_dataAr + dev_dataAr + test_dataAr random.shuffle(train_dev_dataAr) train_dataAr = train_dev_dataAr[:int(len(train_dev_dataAr) * 0.9)] dev_dataAr = train_dev_dataAr[int(len(train_dev_dataAr) * 0.9):] print('Spanish translated data:') trainfileEs = 'translated/Es_' + emotion + '_train.txt' devfileEs = 'translated/Es_' + emotion + '_dev.txt'