import torch import numpy as np import torch.nn as nn from torch.autograd import Variable import matplotlib.pyplot as plt class RNN(nn.Module): def __init__(self): super(RNN, self).__init__() self.rnn = nn.RNN( input_size=1, hidden_size=32, num_layers=1, batch_first=True ) self.out = nn.Linear(32, 1) def forward(self, x, h_state): # shape # 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) outs = [] for time_step in range(r_out.size(1)): outs.append(self.out(r_out[:, time_step, :])) return torch.stack(outs, dim=1), h_state rnn = RNN().cuda() optimizer = torch.optim.Adam(rnn.parameters(), lr=0.01) loss_func = nn.MSELoss() plt.ion() plt.show() plt.figure(figsize=(12,6)) h_state = None for step in range(50): start, end = step*np.pi, (step+1)*np.pi steps = np.linspace(start, end, 10, dtype=np.float32) x_np = np.sin(steps) y_np = np.cos(steps) x = Variable(torch.from_numpy(x_np[np.newaxis, :, np.newaxis])).cuda() # shape (batch, time_step, input_size) y = Variable(torch.from_numpy(y_np[np.newaxis, :, np.newaxis])).cuda() prediction, h_state = rnn(x, h_state) h_state = Variable(h_state.data).cuda() loss = loss_func(prediction, y) optimizer.zero_grad() loss.backward() optimizer.step() print('loss=%.2f' % loss) plt.plot(steps, y.cpu().data[0], 'r-', lw=1) plt.plot(steps, prediction.cpu().data[0], 'b-', lw=1) plt.pause(0.2) plt.ioff() plt.show()
结果：
loss=0.58 loss=0.53 loss=0.55 ... loss=0.03 loss=0.01 loss=0.00
