from datasets.conll2003 import Conll2003Dataset
from models.bilstm_crf.bilstm_crf import BilstmCRF
if __name__ == "__main__":
device = torch.device("cpu" if not torch.cuda.is_available() else "cuda")
batch_size = 200
print(f"Device: {device}")
models_dir = 'experiments/models/saves/'
# ----------------------------------- Prepare dataset ------------------------------------
TASK_TYPE = 'NER'
LABELED_UNLABELED_RATIO = 0.1
ID_sim_beta = 1
dataset = Conll2003Dataset(
save_file_path='datasets/saves/conll2003NER.pkl', task='NER')
train_dataset, test_dataset = dataset.train_test_split()
ltrain_dataset, utrain_dataset = train_dataset.split(
LABELED_UNLABELED_RATIO)
labeled_train_x, labeled_train_y = ltrain_dataset.x_embeddings, ltrain_dataset.y
unlabeled_train_x, unlabeled_train_y = utrain_dataset.x_embeddings, utrain_dataset.y
test_dataset_x, test_dataset_y = test_dataset.x_embeddings, test_dataset.y
print(f"labelled size: {len(labeled_train_y)}",
f"Unlabelled size: {len(unlabeled_train_y)}",
f"test size: {len(test_dataset_y)})")
# ------------------------------------ create model ---------------------------------------
dropout_rate = 0.5
model = BilstmCRF(dataset.max_word_idx + 1,
dataset.max_tag_idx + 1,
300,
from active_learning.utils import active_learning_loop_limited_tokens
from datasets.conll2003 import Conll2003Dataset
from models.supervised.supervised_mc import SupervisedModelAL
if __name__ == "__main__":
device = torch.device("cpu" if not torch.cuda.is_available() else "cuda")
print(f"Device: {device}")
models_dir = 'experiments/models/saves/'
# ----------------------------------- Prepare dataset ------------------------------------
TASK_TYPE = 'NER' if len(sys.argv) < 2 else sys.argv[1]
LABELED_UNLABELED_RATIO = 0.1
ID_sim_beta = 1
train_dataset, test_dataset = Conll2003Dataset(
task=TASK_TYPE).train_test_split()
ltrain_dataset, utrain_dataset = train_dataset.split(
LABELED_UNLABELED_RATIO)
labeled_train_x, labeled_train_y = ltrain_dataset.x_embeddings, ltrain_dataset.y
unlabeled_train_x, unlabeled_train_y = utrain_dataset.x_embeddings, utrain_dataset.y
test_dataset_x, test_dataset_y = test_dataset.x_embeddings, test_dataset.y
print(f"labelled size: {len(labeled_train_y)}",
f"Unlabelled size: {len(unlabeled_train_y)}",
f"test size: {len(test_dataset_y)})")
# ------------------------------------ create model ---------------------------------------
dropout_rate = 0.4
model = SupervisedModelAL(train_dataset.max_tag_idx + 1, 100, 2,
dropout_rate, dropout_rate)
if device.type == 'cuda':
from datasets.conll2003 import Conll2003Dataset
from models.bilstm_crf.bilstm_crf import BilstmCRF
if __name__ == "__main__":
device = torch.device("cpu" if not torch.cuda.is_available() else "cuda")
batch_size = 200
print(f"Device: {device}")
models_dir = 'experiments/models/saves/'
# ----------------------------------- Prepare dataset ------------------------------------
TASK_TYPE = 'POS'
LABELED_UNLABELED_RATIO = 0.1
ID_sim_beta = 1
dataset = Conll2003Dataset(
save_file_path='datasets/saves/conll2003POS.pkl', task='POS')
train_dataset, test_dataset = dataset.train_test_split()
ltrain_dataset, utrain_dataset = train_dataset.split(
LABELED_UNLABELED_RATIO)
labeled_train_x, labeled_train_y = ltrain_dataset.x, ltrain_dataset.y
unlabeled_train_x, unlabeled_train_y = utrain_dataset.x, utrain_dataset.y
test_dataset_x, test_dataset_y = test_dataset.x, test_dataset.y
print(f"labelled size: {len(labeled_train_y)}",
f"Unlabelled size: {len(unlabeled_train_y)}",
f"test size: {len(test_dataset_y)})")
# ------------------------------------ create model ---------------------------------------
dropout_rate = 0.5
model = BilstmCRF(dataset.max_word_idx + 1,
dataset.max_tag_idx + 1,
30,