差分
このページの2つのバージョン間の差分を表示します。
次のリビジョン | 前のリビジョン | ||
pytorch:regression [2022/06/02 13:48] – 作成 watalu | pytorch:regression [2022/06/02 13:56] (現在) – watalu | ||
---|---|---|---|
行 61: | 行 61: | ||
< | < | ||
device = torch.device(" | device = torch.device(" | ||
- | # Assume that we are on a CUDA machine, then this should print a CUDA device: | + | |
print(" | print(" | ||
class Net(torch.nn.Module): | class Net(torch.nn.Module): | ||
def __init__(self, | def __init__(self, | ||
super(Net, self).__init__() | super(Net, self).__init__() | ||
- | self.hidden = torch.nn.Linear(cols, | + | self.hidden = torch.nn.Linear(cols, |
- | self.predict = torch.nn.Linear(size_hidden, | + | self.predict = torch.nn.Linear(size_hidden, |
def forward(self, | def forward(self, | ||
- | x = F.relu(self.hidden(x)) | + | x = F.relu(self.hidden(x)) |
- | x = self.predict(x) | + | x = self.predict(x) |
return x | return x | ||
model = Net(cols, size_hidden, | model = Net(cols, size_hidden, | ||
- | optimizer = torch.optim.Adam(net.parameters(), | + | optimizer = torch.optim.Adam(model.parameters(), |
- | criterion = torch.nn.MSELoss(reduction=' | + | criterion = torch.nn.MSELoss(reduction=' |
</ | </ | ||
行 119: | 行 119: | ||
ax.plot(loss_train_history, | ax.plot(loss_train_history, | ||
ax.plot(loss_test_history, | ax.plot(loss_test_history, | ||
+ | plt.show() | ||
+ | fig, ax = plt.subplots() | ||
+ | ax.plot(loss_train_history, | ||
+ | ax.plot(loss_test_history, | ||
+ | plt.ylim(0, 8000) | ||
+ | plt.show() | ||
+ | plt.plot(loss_test_records) | ||
plt.show() | plt.show() | ||
行 127: | 行 134: | ||
pred=result.data[:, | pred=result.data[:, | ||
print(len(pred), | print(len(pred), | ||
- | r2_score(pred, | + | print(r2_score(pred, |
X = Variable(torch.FloatTensor(X_test)) | X = Variable(torch.FloatTensor(X_test)) | ||
行 133: | 行 140: | ||
pred=result.data[:, | pred=result.data[:, | ||
print(len(pred), | print(len(pred), | ||
- | r2_score(pred, | + | print(r2_score(pred, |
</ | </ | ||