import torch
from torch import nn
import torchvision.datasets as dsets
import torchvision.transforms as transforms
import matplotlib.pyplot as plt

# 超参数
# Hyper Parameters
# 训练整批数据多少次, 为了节约时间, 只训练一次
EPOCH = 1               # train the training data n times, to save time, we just train 1 epoch
BATCH_SIZE = 64
TIME_STEP = 28          # rnn time step / image height    时间步数 / 图片高度
INPUT_SIZE = 28         # rnn input size / image width    每步输入值 / 图片每行像素
LR = 0.01               # learning rate
DOWNLOAD_MNIST = True   # set to True if haven't download the data

# Mnist 手写数字
# Mnist digital dataset
train_data = dsets.MNIST(
    root='./mnist/',                    # 保存或者提取位置
    train=True,                         # this is training data
    transform=transforms.ToTensor(),    # Converts a PIL.Image or numpy.ndarray to
                                        # torch.FloatTensor of shape (C x H x W) and normalize in the range [0.0, 1.0]
    download=DOWNLOAD_MNIST,            # download it if you don't have it
)

# plot one example
print(train_data.train_data.size())     # (60000, 28, 28)
print(train_data.train_labels.size())   # (60000)
plt.imshow(train_data.train_data[0].numpy(), cmap='gray')
plt.title('%i' % train_data.train_labels[0])
plt.show()

# 数据加载
# Data Loader for easy mini-batch return in training  批训练 50samples, 1 channel, 28x28 (50, 1, 28, 28)
train_loader = torch.utils.data.DataLoader(dataset=train_data, batch_size=BATCH_SIZE, shuffle=True)
# 测试数据
# convert test data into Variable, pick 2000 samples to speed up testing
test_data = dsets.MNIST(root='./mnist/', train=False, transform=transforms.ToTensor())
test_x = test_data.test_data.type(torch.FloatTensor)[:2000]/255.   # shape (2000, 28, 28) value in range(0,1)
test_y = test_data.test_labels.numpy()[:2000]    # covert to numpy array

# 用一个 class 来建立 RNN 模型.
# 这个 RNN 整体流程：
# (input0, state0) -> LSTM -> (output0, state1);
# (input1, state1) -> LSTM -> (output1, state2);
# …
# (inputN, stateN)-> LSTM -> (outputN, stateN+1);
# outputN -> Linear -> prediction.
# 通过LSTM分析每一时刻的值, 并且将这一时刻和前面时刻的理解合并在一起, 生成当前时刻对前面数据的理解或记忆. 传递这种理解给下一时刻分析.
# 定义神经网络
class RNN(nn.Module):
    def __init__(self):
        super(RNN, self).__init__()

        self.rnn = nn.LSTM(         # if use nn.RNN(), it hardly learns  LSTM 效果要比 nn.RNN() 好多了
            input_size=INPUT_SIZE,
            hidden_size=64,         # rnn hidden unit
            num_layers=1,           # number of rnn layer  有几层 RNN layers
            batch_first=True,       # input & output will has batch size as 1s dimension. e.g. (batch, time_step, input_size)   input & output 会是以 batch size 为第一维度的特征集 e.g. (batch, time_step, input_size)
        )
        # 输出层
        self.out = nn.Linear(64, 10)

    def forward(self, x):
        # x shape (batch, time_step, input_size)
        # r_out shape (batch, time_step, output_size)
        # h_n shape (n_layers, batch, hidden_size)  # LSTM 有两个 hidden states, h_n 是分线, h_c 是主线
        # h_c shape (n_layers, batch, hidden_size)
        r_out, (h_n, h_c) = self.rnn(x, None)   # None represents zero initial hidden state

        # 选取最后一个时间点的 r_out 输出
        # choose r_out at the last time step
        out = self.out(r_out[:, -1, :]) # 这里 r_out[:, -1, :] 的值也是 h_n 的值
        return out

rnn = RNN()
print(rnn)
# 选择优化器
optimizer = torch.optim.Adam(rnn.parameters(), lr=LR)   # optimize all cnn parameters
# 选择损失函数
loss_func = nn.CrossEntropyLoss()                       # the target label is not one-hotted

# 训练和测试
# 将图片数据看成一个时间上的连续数据, 每一行的像素点都是这个时刻的输入,
# 读完整张图片就是从上而下的读完了每行的像素点. 然后我们就可以拿出 RNN 在最后一步的分析值判断图片是哪一类了.
# training and testing
for epoch in range(EPOCH):
    for step, (b_x, b_y) in enumerate(train_loader):    # gives batch data
        b_x = b_x.view(-1, 28, 28)                      # reshape x to (batch, time_step, input_size)

        output = rnn(b_x)                               # rnn output   # 喂给 rnn net 训练数据 b_x, 输出预测值
        loss = loss_func(output, b_y)                   # cross entropy loss  # 计算两者的误差
        optimizer.zero_grad()                           # clear gradients for this training step  # 清空上一步的残余更新参数值
        loss.backward()                                 # backpropagation, compute gradients      # 误差反向传播, 计算参数更新值
        optimizer.step()                                # apply gradients   # 将参数更新值施加到 rnn net 的 parameters 上

        if step % 50 == 0:
            test_output = rnn(test_x)                   # (samples, time_step, input_size)
            pred_y = torch.max(test_output, 1)[1].data.numpy()
            accuracy = float((pred_y == test_y).astype(int).sum()) / float(test_y.size)
            print('Epoch: ', epoch, '| train loss: %.4f' % loss.data.numpy(), '| test accuracy: %.2f' % accuracy)

# print 10 predictions from test data
test_output = rnn(test_x[:10].view(-1, 28, 28))
pred_y = torch.max(test_output, 1)[1].data.numpy()
print(pred_y, 'prediction number')
print(test_y[:10], 'real number')