model_path = "trained_models/128_hidden/run2/best_model_1fc.pt"
model.load_state_dict(torch.load(model_path))
model.eval()
print("Loaded trained model weights successfully!")
#######################################
################## 2. #################
#######################################
os.makedirs("activations", exist_ok=True)
for i, (data, true_labels) in enumerate(testing_dataloader):
data = data.type(torch.FloatTensor)
true_labels = true_labels.type(torch.LongTensor)
stg_activations, tp_activations, ifg_activations, word_predictions = model.out(
data)
torch.save(stg_activations, 'activations/stg_{}.pt'.format(i))
torch.save(tp_activations, 'activations/tp_{}.pt'.format(i))
torch.save(ifg_activations, 'activations/ifg_{}.pt'.format(i))
'''
print(stg_activations.shape) => torch.Size([2130, 3, 120])
print(tp_activations.shape) => torch.Size([30, 71, 128])
print(ifg_activations.shape) => torch.Size([30, 71, 128])
sys.exit(0)
'''
print(stg_activations.shape)
print(tp_activations.shape)
print(ifg_activations.shape)
best_acc = 0
training_loss = []
test_accuracy = []
for epoch in range(epochs):
total_loss = 0
'''
START EVALUATION PROCEDURE (Start evaluation already before training for comparison)
'''
with torch.no_grad():
correct = 0
total = 0
for i, (data, true_labels) in enumerate(testing_dataloader):
data = data.type(torch.FloatTensor)
true_labels = true_labels.type(torch.LongTensor)
output_conv, output_lstm1, output_lstm1, predictions = model.out(
data)
total += true_labels.size(0)
correct += (predictions.max(1)[1] == true_labels).sum().item()
accuracy = 100 * correct / total
print('Accuracy of the network on the evaluation dataset: %d %%' %
(100 * correct / total))
if accuracy > best_acc:
best_acc == accuracy
# SAVE MODEL
print("SAVING MODEL")
torch.save(model.state_dict(), "trained_models/best_model.pt")
test_accuracy.append(accuracy)
