def run(model, loss, epochs=30):
for epoch in range(epochs):
model.train()
run_epoch(data_generator(V, 8, 20), model, loss)
model.eval()
run_epoch(data_generator(V, 8, 5), model, loss)
# 模型进入测试模式
model.eval()
# 假定的输入张量
source = Variable(torch.LongTensor([[1, 3, 2, 5, 4, 6, 7, 8, 9, 10]]))
# 定义源数据掩码张量, 因为元素都是1, 在我们这里1代表不遮掩
# 因此相当于对源数据没有任何遮掩.
source_mask = Variable(torch.ones(1, 1, 10))
# 最后将model, src, src_mask, 解码的最大长度限制max_len, 默认为10
# 以及起始标志数字, 默认为1, 我们这里使用的也是1
result = greedy_decode(model,
source,
source_mask,
max_len=10,
start_symbol=1)
print(result)
def run(model, loss, V, epochs=10):
start = time.time()
for epoch in range(epochs):
model.train()
# 每批次8个，一个批次循环20次
run_epoch(data_generator(V, 8, 20), model, loss)
model.eval()
print("epochs:%d || time:%s " %(epoch+1, timeSince(start)))
run_epoch(data_generator(V, 8, 5), model, loss)
print("--------------------训练完毕，保存模型-----------------")
torch.save(model.state_dict(), "./model/transform_"+TYPE+".pt")
# 模型进入测试模式
model.eval()
# 假定的输入张量
source = Variable(torch.LongTensor([[1,3,2,5,4,6,7,8,9,10]])).to(device)
# 定义源数据掩码张量, 因为元素都是1, 在我们这里1代表不遮掩
# 因此相当于对源数据没有任何遮掩.
source_mask = Variable(torch.ones(1, 1, 10)).to(device)
# 最后将model, src, src_mask, 解码的最大长度限制max_len, 默认为10
# 以及起始标志数字, 默认为1, 我们这里使用的也是1
result = greedy_decode(model, source, source_mask, max_len=10, start_symbol=1)
print("输入的数据:", source)
print("预测结果是:",result)
def predict(model, sources):
# 模型进入测试模式
# print(sources)
checkpoint = torch.load("./model/transform_"+TYPE+".pt",map_location=device)
model.load_state_dict(checkpoint)
model.eval()
source_mask = Variable(torch.ones(1, 1, 10)).to(device)
num = 1
acc_count = 0
for i in sources:
source = Variable(torch.LongTensor([[int(j) for j in i.strip("\n").strip(" ").split(",")]])).to(device)
result = greedy_decode(model, source, source_mask, max_len=10, start_symbol=1)
if torch.equal(source, result):
acc_count += 1
print("第{} 输入数据为:{}".format(num,source))
print("第{}数据预测的结果是:{}".format(num,result))
num += 1
print("准确率为:",acc_count/ len(sources))
def run(model, loss, epochs=10):
for epoch in range(epochs):
model.train()
run_epoch(data_generator(V, 8, 20), model, loss)
model.eval()
run_epoch(data_generator(V, 8, 5), model, loss)
model.eval()
source = Variable(torch.LongTensor([[1, 3, 2, 5, 4, 6, 7, 8, 9, 10]]))
source_mask = Variable(torch.ones(1, 1, 10))
result = greedy_decode(model,
source,
source_mask,
max_len=10,
start_symbol=1)
print(result)