def init_multitask_autoencoder_learner(
num_features, autoencoder_bottle_neck_feature_size,
autoencoder_num_layers, shared_hidden_layer_size,
user_dense_layer_hidden_size, num_classes, num_covariates,
shared_layer_dropout_prob, user_head_dropout_prob, learning_rate,
decay, class_weights, student_list):
class_weights = torch.tensor(class_weights)
model = multitask_autoencoder.MultiTaskAutoEncoderLearner(
conversions.prepend_ids_with_string(student_list,
"student_"), num_features,
autoencoder_bottle_neck_feature_size, autoencoder_num_layers,
shared_hidden_layer_size, user_dense_layer_hidden_size, num_classes,
num_covariates, shared_layer_dropout_prob, user_head_dropout_prob)
if torch.cuda.is_available():
model.cuda()
class_weights = class_weights.cuda()
reconstruction_criterion = torch.nn.L1Loss(reduction="sum")
classification_criterion = torch.nn.CrossEntropyLoss(weight=class_weights)
optimizer = torch.optim.Adam(model.parameters(),
lr=learning_rate,
weight_decay=decay)
return model, reconstruction_criterion, classification_criterion, optimizer
for split_no, split in enumerate(splits):
best_split_score = -1
epoch_at_best_score = 0
tensorified_data['train_ids'] = split['train_ids']
data['train_ids'] = split['train_ids']
tensorified_data['val_ids'] = split['val_ids']
data['val_ids'] = split['val_ids']
tensorified_data['test_ids'] = []
validation_user_statistics_over_epochs = []
model = multitask_autoencoder.MultiTaskAutoEncoderLearner(
conversions.prepend_ids_with_string(student_list, "student_"),
num_features, autoencoder_bottle_neck_feature_size,
autoencoder_num_layers, shared_hidden_layer_size,
user_dense_layer_hidden_size, num_classes, num_covariates,
shared_layer_dropout_prob, user_head_dropout_prob)
if cuda_enabled:
model.cuda()
class_weights = class_weights.cuda()
reconstruction_criterion = torch.nn.L1Loss(reduction="sum")
classification_criterion = torch.nn.CrossEntropyLoss(
weight=class_weights)
optimizer = torch.optim.Adam(model.parameters(),
lr=learning_rate,
weight_decay=decay)