Author: chen_h
Micro Signal & QQ: 862251340
micro-channel public number: coderpai

(E) pytorch study notes

RNN Recurrent Neural Network (Classification)

Recurrent neural network so that the neural network with the memory, the data sequence of words, recurrent neural network can achieve better results. If you have no particular knowledge of the recurrent neural network, please watch the short animation a few minutes, RNN animated introduction and LSTM About animation allows you to vividly understand RNN. then we do a step by step analysis of handwritten digits RNN it.

MNIST handwritten data

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


torch.manual_seed(1)    # reproducible

# Hyper Parameters
EPOCH = 1           # 训练整批数据多少次, 为了节约时间, 我们只训练一次
BATCH_SIZE = 64
TIME_STEP = 28      # rnn 时间步数 / 图片高度
INPUT_SIZE = 28     # rnn 每步输入值 / 图片每行像素
LR = 0.01           # learning rate
DOWNLOAD_MNIST = True  # 如果你已经下载好了mnist数据就写上 Fasle


# Mnist 手写数字
train_data = torchvision.datasets.MNIST(
    root='./mnist/',    # 保存或者提取位置
    train=True,  # this is training data
    transform=torchvision.transforms.ToTensor(),    # 转换 PIL.Image or numpy.ndarray 成
                                                    # torch.FloatTensor (C x H x W), 训练的时候 normalize 成 [0.0, 1.0] 区间
    download=DOWNLOAD_MNIST,          # 没下载就下载, 下载了就不用再下了
)

Here Insert Picture Description

The value of the local black is 0 and white where the value is greater than zero.

Also, apart from the training data, and gave some test data, test to see if it does not have good training.

test_data = torchvision.datasets.MNIST(root='./mnist/', train=False)

# 批训练 50samples, 1 channel, 28x28 (50, 1, 28, 28)
train_loader = Data.DataLoader(dataset=train_data, batch_size=BATCH_SIZE, shuffle=True)

# 为了节约时间, 我们测试时只测试前2000个
test_x = torch.unsqueeze(test_data.test_data, dim=1).type(torch.FloatTensor)[:2000]/255.   # shape from (2000, 28, 28) to (2000, 1, 28, 28), value in range(0,1)
test_y = test_data.test_labels[:2000]

RNN model

As before, we use a class to create RNN model. This whole process is RNN

(input0, state0) -> LSTM -> (output0, state1);
(input1, state1) -> LSTM -> (output1, state2);
…
(inputN, stateN)-> LSTM -> (outputN, stateN+1);
outputN-> Linear-> predictionBy LSTManalysis value every moment, and it will be appreciated that the time and the preceding time point were combined together, generate the current time or memory understanding of the foregoing data transmitted to the next time this understanding analysis.

class RNN(nn.Module):
    def __init__(self):
        super(RNN, self).__init__()

        self.rnn = nn.LSTM(     # LSTM 效果要比 nn.RNN() 好多了
            input_size=28,      # 图片每行的数据像素点
            hidden_size=64,     # rnn hidden unit
            num_layers=1,       # 有几层 RNN layers
            batch_first=True,   # input & output 会是以 batch size 为第一维度的特征集 e.g. (batch, time_step, input_size)
        )

        self.out = nn.Linear(64, 10)    # 输出层

    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 表示 hidden state 会用全0的 state

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

rnn = RNN()
print(rnn)
"""
RNN (
  (rnn): LSTM(28, 64, batch_first=True)
  (out): Linear (64 -> 10)
)
"""

training

We will picture data as continuous data on a time, each row of pixels is input this time, the entire image is read from the top down to read the pixels per row. Then we can come up with RNN in the final step of the analysis to determine the value of what kind of picture. the following code omitted from the calculation accuracysection, you can see all the code in my github.

optimizer = torch.optim.Adam(rnn.parameters(), lr=LR)   # optimize all parameters
loss_func = nn.CrossEntropyLoss()   # the target label is not one-hotted

# training and testing
for epoch in range(EPOCH):
    for step, (x, b_y) in enumerate(train_loader):   # gives batch data
        b_x = x.view(-1, 28, 28)   # reshape x to (batch, time_step, input_size)

        output = rnn(b_x)               # rnn output
        loss = loss_func(output, b_y)   # cross entropy loss
        optimizer.zero_grad()           # clear gradients for this training step
        loss.backward()                 # backpropagation, compute gradients
        optimizer.step()                # apply gradients
"""
...
Epoch:  0 | train loss: 0.0945 | test accuracy: 0.94
Epoch:  0 | train loss: 0.0984 | test accuracy: 0.94
Epoch:  0 | train loss: 0.0332 | test accuracy: 0.95
Epoch:  0 | train loss: 0.1868 | test accuracy: 0.96
"""

Finally, we come to take 10 data to see predicted value right in the end:

test_output = rnn(test_x[:10].view(-1, 28, 28))
pred_y = torch.max(test_output, 1)[1].data.numpy().squeeze()
print(pred_y, 'prediction number')
print(test_y[:10], 'real number')
"""
[7 2 1 0 4 1 4 9 5 9] prediction number
[7 2 1 0 4 1 4 9 5 9] real number
"""

RNN Recurrent Neural Networks (return)

Recurrent neural network so that the neural network with the memory, the data sequence of words, recurrent neural network can achieve better results. If you have no particular knowledge of the recurrent neural network, please watch the short animation a few minutes, RNN animated introduction and LSTM About animation allows you to vividly understand RNN. the last time we saw earlier with reference to the last time point output RNN to determine which type picture, this time we really come up with a timely RNN to predict the time series.

Here Insert Picture Description

Training data

We use the data of some of the data is such that we want to use sinthe curve predicted coscurve.

Here Insert Picture Description

import torch
from torch import nn
import numpy as np
import matplotlib.pyplot as plt

torch.manual_seed(1)    # reproducible

# Hyper Parameters
TIME_STEP = 10      # rnn time step / image height
INPUT_SIZE = 1      # rnn input size / image width
LR = 0.02           # learning rate
DOWNLOAD_MNIST = False  # set to True if haven't download the data

RNN model

This time the RNN, we each r_outhad to put Linearto the calculated predicted output, so we can use a for loop to calculate the loop. This is Tensorflow unmatched! In addition to this, there are some dynamic process can be in this tutorial view, a look at our PyTorch and Tensorflow Which in the end strong.

class RNN(nn.Module):
    def __init__(self):
        super(RNN, self).__init__()

        self.rnn = nn.RNN(  # 这回一个普通的 RNN 就能胜任
            input_size=1,
            hidden_size=32,     # rnn hidden unit
            num_layers=1,       # 有几层 RNN layers
            batch_first=True,   # input & output 会是以 batch size 为第一维度的特征集 e.g. (batch, time_step, input_size)
        )
        self.out = nn.Linear(32, 1)

    def forward(self, x, h_state):  # 因为 hidden state 是连续的, 所以我们要一直传递这一个 state
        # x (batch, time_step, input_size)
        # h_state (n_layers, batch, hidden_size)
        # r_out (batch, time_step, output_size)
        r_out, h_state = self.rnn(x, h_state)   # h_state 也要作为 RNN 的一个输入

        outs = []    # 保存所有时间点的预测值
        for time_step in range(r_out.size(1)):    # 对每一个时间点计算 output
            outs.append(self.out(r_out[:, time_step, :]))
        return torch.stack(outs, dim=1), h_state


rnn = RNN()
print(rnn)
"""
RNN (
  (rnn): RNN(1, 32, batch_first=True)
  (out): Linear (32 -> 1)
)
"""

In fact RNN familiar friends should know that forwardthe process of seeking the point of output for each time there is a move to make the calculation relatively small amount, but above content is primarily in order to present PyTorch advantage in the dynamic composition, so I used a for loopto building set of output system. the following describes an alternative embodiment. reshape using the calculated bulk manner.

def forward(self, x, h_state):
    r_out, h_state = self.rnn(x, h_state)
    r_out = r_out.view(-1, 32)
    outs = self.out(r_out)
    return outs.view(-1, 32, TIME_STEP), h_state

Training ¶

The following code can achieve a motion picture effect friends ~ fun, it can be seen that we use xas an input sinvalue and yan output fitting desired, cosvalue. Because they are the two curves of the existence of a relationship, we be able to use sinto predict cos. rnnwill understand their relationship, and to analyze it with this time inside the parameters of sinhow the point on the curve corresponds to the cospoint on the curve.

Here Insert Picture Description

optimizer = torch.optim.Adam(rnn.parameters(), lr=LR)   # optimize all rnn parameters
loss_func = nn.MSELoss()

h_state = None   # 要使用初始 hidden state, 可以设成 None

for step in range(100):
    start, end = step * np.pi, (step+1)*np.pi   # time steps
    # sin 预测 cos
    steps = np.linspace(start, end, 10, dtype=np.float32)
    x_np = np.sin(steps)    # float32 for converting torch FloatTensor
    y_np = np.cos(steps)

    x = torch.from_numpy(x_np[np.newaxis, :, np.newaxis])    # shape (batch, time_step, input_size)
    y = torch.from_numpy(y_np[np.newaxis, :, np.newaxis])

    prediction, h_state = rnn(x, h_state)   # rnn 对于每个 step 的 prediction, 还有最后一个 step 的 h_state
    # !!  下一步十分重要 !!
    h_state = h_state.data  # 要把 h_state 重新包装一下才能放入下一个 iteration, 不然会报错

    loss = loss_func(prediction, y)     # cross entropy loss
    optimizer.zero_grad()               # clear gradients for this training step
    loss.backward()                     # backpropagation, compute gradients
    optimizer.step()                    # apply gradients