##########################################################################
task_name = args.task_name.lower()
if task_name not in processors and 'race' not in task_name:
raise ValueError("Task not found: %s" % (task_name))
if 'race' in task_name:
pass
else:
processor = processors[task_name]()
output_mode = output_modes[task_name]
label_list = processor.get_labels()
num_labels = len(label_list)
tokenizer = BertTokenizer.from_pretrained(args.bert_model, do_lower_case=True)
if args.do_train:
train_sampler = SequentialSampler if DEBUG else RandomSampler
read_set = 'train'
if args.teacher_prediction is not None and args.alpha > 0:
logger.info('loading teacher\'s predictoin')
teacher_predictions = pickle.load(open(
args.teacher_prediction,
'rb'))['train'] if args.teacher_prediction is not None else None
logger.info(
'teacher acc = %.2f, teacher loss = %.5f' %
(teacher_predictions['acc'] * 100, teacher_predictions['loss']))
if args.kd_model == 'kd':
train_examples, train_dataloader, _ = get_task_dataloader(
task_name,
# We convert the raw MRC data into torch tensors so we don't have to convert it everytime we need it
import pickle
import os
from src.race_data_processing import read_mrc_examples, convert_examples_to_features
from envs import HOME_DATA_FOLDER
from BERT.pytorch_pretrained_bert.tokenization import BertTokenizer
tokenizer = BertTokenizer.from_pretrained(os.path.join(
HOME_DATA_FOLDER, 'models/pretrained/bert-base-uncased'),
do_lower_case=True)
data_dir = os.path.join(HOME_DATA_FOLDER, 'data_raw')
data_output_dir = os.path.join(HOME_DATA_FOLDER, 'data_feat')
# for task_name in ['race-high', 'race-merge', 'race-middle']:
for task_name in ['race-merge', 'race-high', 'race-middle']:
print(task_name, end='...\t')
output_folder = os.path.join(data_output_dir, task_name)
if not os.path.exists(output_folder):
os.mkdir(output_folder)
for s in ['train', 'dev', 'test']:
train_examples = read_mrc_examples(
os.path.join(data_dir, task_name, s + '.' + task_name + '.json'))
features = convert_examples_to_features(train_examples, tokenizer, 512,
True)
with open(os.path.join(output_folder, s + '.' + task_name + '.pkl'),
'wb') as f:
pickle.dump(features, f)