Created
January 4, 2018 07:20
-
-
Save soumith/ceb0d3de23585e676fd3b5e0402a45a3 to your computer and use it in GitHub Desktop.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
import torch | |
import torch.nn as nn | |
from torch.autograd import Variable | |
import numpy as np | |
BATCH_SIZE = 1 | |
INPUT_DIM = 1 | |
OUTPUT_DIM = 1 | |
DTYPE = np.float32 | |
class Net(nn.Module): | |
def __init__(self, input_dim, hidden_dim, output_dim, hidden_layers): | |
super(Net, self).__init__() | |
self.input_dim = input_dim | |
self.hidden_dim = hidden_dim | |
self.output_dim = output_dim | |
self.hidden_layers = hidden_layers | |
self.rnn = nn.RNN(input_dim, hidden_dim, hidden_layers) | |
self.h2o = nn.Linear(hidden_dim, output_dim) | |
def forward(self, x): | |
h_0 = Variable(torch.zeros(self.hidden_layers, BATCH_SIZE, self.hidden_dim)) | |
if DTYPE == np.float32: | |
h_0.float() | |
else: | |
h_0.double() | |
output, h_t = self.rnn(x, h_0) | |
output = self.h2o(output) | |
return output | |
def weights_init(m): | |
if isinstance(m, nn.RNN): | |
nn.init.xavier_uniform(m.weight_ih_l0.data) | |
nn.init.orthogonal(m.weight_hh_l0.data) | |
nn.init.constant(m.bias_ih_l0.data, 0) | |
nn.init.constant(m.bias_hh_l0.data, 0) | |
if isinstance(m, nn.Linear): | |
nn.init.xavier_uniform(m.weight.data) | |
nn.init.constant(m.bias.data, 0) | |
data = np.loadtxt('mg17.csv', delimiter=',', dtype=DTYPE) | |
trX = torch.from_numpy(np.expand_dims(data[:4000, [0]], axis=1)) | |
trY = torch.from_numpy(np.expand_dims(data[:4000, [1]], axis=1)) | |
loss_fcn = nn.MSELoss() | |
model = Net(INPUT_DIM, 10, OUTPUT_DIM, 1) | |
model.apply(weights_init) | |
optimizer = torch.optim.Adam(model.parameters(), lr=0.01) | |
for e in range(1000): | |
model.train() | |
x = Variable(trX) | |
y = Variable(trY) | |
model.zero_grad() | |
output = model(x) | |
loss = loss_fcn(output, y) | |
loss.backward() | |
optimizer.step() | |
print("Epoch", e + 1, "TR:", loss.cpu().data.numpy()[0]) | |
y = Variable(trY) | |
model.zero_grad() | |
output = model(x) | |
loss = loss_fcn(output, y) | |
loss.backward() | |
optimizer.step() | |
print("Epoch", e + 1, "TR:", loss.cpu().data.numpy()[0]) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment