def train_model(
self,
dataset_handler: inlp_dataset_handler.ClassificationDatasetHandler
) -> float:
"""
:param dataset_handler:
:return: accuracy score on the dev set / Person's R in the case of regression
"""
X_train, Y_train = dataset_handler.get_current_training_set()
X_train1, X_train2 = X_train
X_dev, Y_dev = dataset_handler.get_current_dev_set()
X_dev1, X_dev2 = X_dev
device = self.model_params["device"]
train_dataset = Dataset(X_train1, X_train2, Y_train, device=device)
dev_dataset = Dataset(X_dev1, X_dev2, Y_dev, device=device)
self.model = self.model_class(
train_dataset,
dev_dataset,
input_dim=self.model_params["input_dim"],
hidden_dim=self.model_params["hidden_dim"],
batch_size=self.model_params["batch_size"],
verbose=self.model_params["verbose"],
same_weights=self.model_params["same_weights"],
compare_by=self.model_params["compare_by"]).to(device)
score = self.model.train_network(self.model_params["num_iter"])
return score
def train_model(self, dataset_handler: inlp_dataset_handler.ClassificationDatasetHandler) -> float:
"""
:param dataset_handler:
:return: accuracy score on the dev set / Person's R in the case of regression
"""
X_train, Y_train = dataset_handler.get_current_training_set()
X_dev, Y_dev = dataset_handler.get_current_dev_set()
self.model.fit(X_train, Y_train)
score = self.model.score(X_dev, Y_dev)
return score
def train_model(
self,
dataset_handler: inlp_dataset_handler.ClassificationDatasetHandler
) -> float:
"""
:param dataset_handler:
:return: accuracy score on the dev set / Person's R in the case of regression
"""
X_train, sents_train, ids_train = dataset_handler.get_current_training_set(
)
X_train1, X_train2 = X_train
X_dev, sents_dev, ids_dev = dataset_handler.get_current_dev_set()
X_dev1, X_dev2 = X_dev
device = self.model_params["device"]
ids_train1, ids_train2 = ids_train
ids_dev1, ids_dev2 = ids_dev
sents_train1, sents_train2 = sents_train
sents_dev1, sents_dev2 = sents_dev
train_dataset = MetricLearningDataset(X_train1,
X_train2,
sents_train1,
sents_train2,
ids_train1,
ids_train2,
device=device)
dev_dataset = MetricLearningDataset(X_dev1,
X_dev2,
sents_dev1,
sents_dev2,
ids_dev1,
ids_dev2,
device=device)
self.model = self.model_class(
train_dataset,
dev_dataset,
input_dim=self.model_params["input_dim"],
hidden_dim=self.model_params["hidden_dim"],
batch_size=self.model_params["batch_size"],
verbose=self.model_params["verbose"],
k=self.model_params["k"],
p=self.model_params["p"],
mode=self.model_params["mode"],
final=self.model_params["final"],
device=self.model_params["device"])
self.model = self.model.to(device)
score = self.model.train_network(self.model_params["num_iter"])
return score
def train_model(
self,
dataset_handler: inlp_dataset_handler.ClassificationDatasetHandler
) -> float:
"""
:param dataset_handler:
:return: accuracy score on the dev set / Person's R in the case of regression
"""
X_train, Y_train = dataset_handler.get_current_training_set()
X_dev, Y_dev = dataset_handler.get_current_dev_set()
train_loader = DataLoader(list(zip(X_train, Y_train)), batch_size=16)
dev_loader = DataLoader(list(zip(X_dev, Y_dev)), batch_size=16)
trainer = pl.Trainer(max_epochs=self.num_epochs,
min_epochs=1,
gpus=1 if self.device_to_use == "cuda" else 0)
trainer.fit(self, train_loader, dev_loader)
score = self.score(X_dev, Y_dev)
return score