f.write('pairID,gold_label' + '\n')
for item in output:
f.write(str(item[0]) + ',' + target_names[item[1]] + '\n')
else:
return 100. * n_correct / n_total, 100. * f1_score(
np.asarray(y_true), np.asarray(y_preds),
average='weighted'), clf_report
if __name__ == "__main__":
dictionary = helper.load_object(args.save_path + 'dictionary.p')
embeddings_index = helper.load_word_embeddings(args.word_vectors_directory,
args.word_vectors_file,
dictionary.word2idx)
class_distributions = None
model = BCN(dictionary, embeddings_index, class_distributions, args)
if args.cuda and torch.cuda.is_available():
torch.cuda.set_device(args.gpu)
model = model.cuda()
helper.load_model_states_from_checkpoint(
model, args.save_path + 'model_best.pth.tar', 'state_dict', args.cuda)
print('vocabulary size = ', len(dictionary))
task_names = ['snli', 'multinli'] if args.task == 'allnli' else [args.task]
for task in task_names:
test_corpus = data.Corpus(args.tokenize)
if 'imdb' in args.task:
###############################################################################
# Load Learning to Skim paper's Pickle file
###############################################################################
train_d, dev_d, test_d = helper.get_splited_imdb_data(
args.dropout_ratio, # float. Used by McCann et al.: 0.1, 0.2 or 0.3
'maxout_reduction':
args.maxout_reduction, # int. Used by McCann et al.: 2, 4 or 8
'bn_decay': args.bn_decay, # float
'bn_epsilon': args.bn_epsilon, # float
'optimizer': args.
optimizer, # "adam" or "gradientdescent". Used by McCann et al.: "adam"
'learning_rate': args.learning_rate, # float. Used by McCann et al.: 0.001
'adam_beta1': args.adam_beta1, # float (used only if optimizer == "adam")
'adam_beta2': args.adam_beta2, # float (used only if optimizer == "adam")
'adam_epsilon':
args.adam_epsilon # float (used only if optimizer == "adam")
}
if args.mode == 1:
BCN(hyperparameters, 3, 128, 900, args.outputdir).dry_run()
sys.exit()
if not os.path.exists(args.outputdir):
os.makedirs(args.outputdir)
if not os.path.exists(os.path.join(args.outputdir, "info.txt")):
with open(os.path.join(args.outputdir, "info.txt"), "w") as outputfile:
outputfile.write(str(args.type) + "\n")
outputfile.write(str(args.transfer_task) + "\n")
outputfile.write(str(hyperparameters))
"""
DATASET
"""
if args.type == "GloVe":
if outfile:
target_names = ['entailment', 'neutral', 'contradiction']
with open(outfile, 'w') as f:
f.write('pairID,gold_label' + '\n')
for item in output:
f.write(str(item[0]) + ',' + target_names[item[1]] + '\n')
else:
return 100. * n_correct / n_total, 100. * f1_score(numpy.asarray(y_true), numpy.asarray(y_preds),
average='weighted')
if __name__ == "__main__":
dictionary = helper.load_object(args.save_path + 'dictionary.p')
embeddings_index = helper.load_word_embeddings(args.word_vectors_directory, args.word_vectors_file,
dictionary.word2idx)
model = BCN(dictionary, embeddings_index, args)
if args.cuda:
torch.cuda.set_device(args.gpu)
model = model.cuda()
helper.load_model_states_from_checkpoint(model, args.save_path + 'model_best.pth.tar', 'state_dict', args.cuda)
print('vocabulary size = ', len(dictionary))
task_names = ['snli', 'multinli'] if args.task == 'allnli' else [args.task]
for task in task_names:
test_corpus = data.Corpus(args.tokenize)
if 'IMDB' in args.task:
###############################################################################
# Load Learning to Skim paper's Pickle file
###############################################################################
train_d, dev_d, test_d = helper.get_splited_imdb_data(args.save_path+'data/'+'imdb.p')
test_corpus.parse(test_d, task, args.max_example)
dictionary = data.Dictionary()
dictionary.build_dict(train_corpus.data + dev_corpus.data + test_corpus.data, args.max_words)
helper.save_object(dictionary, args.output_base_path + args.task+'/' + 'dictionary.p')
print('vocabulary size = ', len(dictionary))
embeddings_index = helper.load_word_embeddings(args.word_vectors_directory, args.word_vectors_file, dictionary.word2idx)
print('number of OOV words = ', len(dictionary) - len(embeddings_index))
# ###############################################################################
# # Build the model
# ###############################################################################
model = BCN(dictionary, embeddings_index, args)
selector = Selector(dictionary, embeddings_index, args)
print (selector)
print (model)
optim_fn_selector, optim_params_selector = helper.get_optimizer(args.optimizer)
optimizer_selector = optim_fn_selector(filter(lambda p: p.requires_grad, selector.parameters()), **optim_params_selector)
optim_fn, optim_params = helper.get_optimizer(args.optimizer)
optimizer = optim_fn(filter(lambda p: p.requires_grad, model.parameters()), **optim_params)
best_acc = 0
param_dict_selector = helper.count_parameters(selector)
param_dict = helper.count_parameters(model)
print('number of trainable parameters = ', numpy.sum(list(param_dict_selector.values())), numpy.sum(list(param_dict.values())), numpy.sum(list(param_dict.values())) + numpy.sum(list(param_dict_selector.values())) )