with open(file_path, 'r', encoding='utf-8') as f:
data = [line.split() for line in f.readlines()]
except:
with open(file_path, 'r', encoding='utf-8') as f:
data = [line.split() for line in f.readlines()]
sentences.extend(data)
labels.extend([label] * len(data))
label += 1
print("\n--------Building essays completed!----------\n")
essays = Eassy.Eassy(sentences, labels)
essays.E2V()
print("\n--------Calculating essay vectors completed!----------\n")
train_x, test_x, train_y, test_y = essays.split()
rnn = Architecture.Representation(64, 64, 64)
# loss function
loss_fun = torch.nn.CrossEntropyLoss()
opt = torch.optim.Adam(rnn.parameters(), lr=0.01)
print("\n--------Starting train!----------\n")
acclist = []
for epoch in range(200):
lab = []
pred = []
for essay, label in zip(train_x, train_y):
essay = torch.unsqueeze(essay, 0)
essay = torch.unsqueeze(essay, 0)
essay = essay.float()
label = torch.tensor([label])
acc = Acc(pred,label)
accList[step].append(acc)
print("\n<<-----epoch:{0} batch:{1} | batch_x_size:{2} | loss:{3:.3f} | acc:{4:.3f}----->>\n".format(epoch,step,batch_x.size(),loss,acc))
loss.backward()
opt.step()
plt.plot(range(1,iter+1),accList[0],marker='*',mec='b',mfc='w')
plt.plot(range(1,iter+1),accList[1],marker='s',mec='g',mfc='w')
plt.xlabel("epoch")
plt.ylabel("acc")
plt.legend()
plt.show()
if __name__ == "__main__":
train_loader,dataLen_train,test_loader,dataLen_test = load_data()
print("dataLen: ",dataLen_train)
model = Architecture.Representation(5, dataLen_train)
loss_fun = torch.nn.CrossEntropyLoss()
opt = torch.optim.Adam(model.parameters(), lr=0.001)
train(train_loader,300)
for step, (batch_x, batch_y) in enumerate(test_loader):
test = model.eval()
out = test(test_loader)
out = out.view(-1, out.shape[2])
# print(out)
pred = torch.max(out, 1)[1].data.numpy()
print(pred)
batch_y = batch_y.view(-1)
loss = loss_fun(out, batch_y)
label = numpy.asarray(batch_y, dtype=int, order=None)
print(label)
