def train_nn_wavg_pr_model(embedding_method, dataset_id, layer, device_id):
data = load_train_data(dataset_id, 'classification')
train, val = stratified_sampling(data)
print(len(train), len(val))
trainer = ModelTrainer(embedding_method, dataset_id, layer, device_id)
model = trainer.train_nn_wavg_pr_model(train, val)
save_model(embedding_method, dataset_id, layer, 'nn_wavg_pr_model', model)
def train_nn_rouge_reg_model(embedding_method, dataset_id, layer, device_id):
data = load_train_data(dataset_id, 'regression_rouge')
train, val = stratified_sampling(data)
print(len(train), len(val))
trainer = ModelTrainer(embedding_method, dataset_id, layer, device_id)
model = trainer.train_nn_rouge_reg_model(train, val)
save_model(embedding_method, dataset_id, layer, 'nn_rouge_reg_model',
model)
def train_nn_sinkhorn_pr_model(embedding_method, dataset_id, layer, device_id):
data = load_train_data(dataset_id, 'classification')
for i, train, val in cross_validation_sampling(data):
print(len(train), len(val))
print(f'Model {i + 1}')
trainer = ModelTrainer(embedding_method, dataset_id, layer, device_id)
model = trainer.train_nn_sinkhorn_pr_model(train, val)
save_model(embedding_method, dataset_id, layer,
f'nn_sinkhorn_pr_model_{i}', model)
def train_cond_nn_wavg_pr_model(embedding_method, dataset_id, layer,
device_id):
data = load_train_data(dataset_id, 'classification')
for i, train, val in cross_validation_sampling(data):
print(len(train), len(val))
print(f'Model {i + 1}')
trainer = ModelTrainer(embedding_method, dataset_id, layer, device_id)
model = trainer.train_cond_nn_wavg_pr_model(train, val)
save_model(embedding_method, dataset_id, layer,
f'cond_nn_wavg_pr_model_{i}', model)
import torch
torch.cuda.empty_cache()