for row in dev_preds:
row = row.tolist()
final_dev_predictions.append(int(max(set(row), key=row.count)))
dev['predictions'] = final_dev_predictions
final_test_predictions = []
else:
model = ClassificationModel(MODEL_TYPE,
MODEL_NAME,
args=args,
use_cuda=torch.cuda.is_available())
model.train_model(train,
macro_f1=macro_f1,
weighted_f1=weighted_f1,
accuracy=sklearn.metrics.accuracy_score)
dev_predictions, raw_dev_outputs = model.predict(dev_sentences)
dev['predictions'] = dev_predictions
dev['predictions'] = decode(dev['predictions'])
dev['labels'] = decode(dev['labels'])
time.sleep(5)
print_information(dev, "predictions", "labels")
dev.to_csv(os.path.join(TEMP_DIRECTORY, "level_2_pred.tsv"),
header=True,
sep='\t',
index=False,
encoding='utf-8')
use_cuda=torch.cuda.is_available())
predictions, raw_outputs = model.predict(test_sentences)
test_preds[:, i] = predictions
print("Completed Fold {}".format(i))
# select majority class of each instance (row)
final_predictions = []
for row in test_preds:
row = row.tolist()
final_predictions.append(int(max(set(row), key=row.count)))
test['predictions'] = final_predictions
else:
model.train_model(train,
macro_f1=macro_f1,
weighted_f1=weighted_f1,
accuracy=sklearn.metrics.accuracy_score)
predictions, raw_outputs = model.predict(test_sentences)
test['predictions'] = predictions
test['predictions'] = decode(test['predictions'])
test['labels'] = decode(test['labels'])
time.sleep(5)
print_information(test, "predictions", "labels")
test.to_csv(os.path.join(TEMP_DIRECTORY, RESULT_FILE),
header=True,
sep='\t',
index=False,
encoding='utf-8')
