import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
from torchvision import datasets, transforms
import numpy as np
import matplotlib.pyplot as plt
from tqdm import tqdm
#データ前処理 transform を設定
transform = transforms.Compose(
[transforms.ToTensor(), # Tensor変換とshape変換 [H, W, C] -> [C, H, W]
transforms.Normalize((0.5, ), (0.5, ))]) # 標準化 平均:0.5 標準偏差:0.5
#訓練用Datasetを作成
train_dataset = datasets.MNIST(root='./data',
train=True,
download=True,
transform=transform)
#検証用Datasetを作成
val_dataset = datasets.MNIST(root='./data',
train=False,
download=True,
transform=transform)
#訓練用 Dataloder
train_dataloader = torch.utils.data.DataLoader(train_dataset,
batch_size=64,
shuffle=True)
#検証用 Dataloder
val_dataloader = torch.utils.data.DataLoader(val_dataset,
batch_size=64,
shuffle=False)
# 辞書型変数にまとめる
dataloaders_dict = {"train": train_dataloader, "val": val_dataloader}
batch_iterator = iter(dataloaders_dict["train"]) # イテレータに変換
imges, labels = next(batch_iterator) # 1番目の要素を取り出す
print("imges size = ", imges.size())
print("labels size = ", labels.size())
#試しに1枚 plot してみる
plt.imshow(imges[0].numpy().reshape(28,28), cmap='gray')
plt.title("label = {}".format(labels[0].numpy()))
plt.show()
imges size = torch.Size([64, 1, 28, 28]) labels size = torch.Size([64])
# 畳み込み層+全結合層のネットワークモデル
class Net(nn.Module):
def __init__(self):
super(Net, self).__init__()
self.conv1 = nn.Conv2d(1, 32, 3, 1) #畳み込み層
self.conv2 = nn.Conv2d(32, 64, 3, 1) #畳み込み層
self.fc1 = nn.Linear(9216, 128) #全結合層
self.fc2 = nn.Linear(128, 10) #全結合層
def forward(self, x):
x = self.conv1(x) # (Batch, 1, 28, 28) -> (Batch, 32, 26, 26)
x = F.relu(x)
x = self.conv2(x) # (Batch, 32, 26, 26) -> (Batch, 64, 24, 24)
x = F.relu(x)
x = F.max_pool2d(x, 2) # (Batch, 64, 24, 24) -> (Batch, 64, 12, 12)
x = torch.flatten(x, 1) # (Batch, 64, 12, 12) -> (Batch, 9216)
x = self.fc1(x) # (Batch, 9216) -> (Batch, 128)
x = self.fc2(x) # (Batch, 128) -> (Batch, 10)
return x
#モデル作成
net = Net()
#ネットワークのレイヤー確認
print(net)
Net( (conv1): Conv2d(1, 32, kernel_size=(3, 3), stride=(1, 1)) (conv2): Conv2d(32, 64, kernel_size=(3, 3), stride=(1, 1)) (fc1): Linear(in_features=9216, out_features=128, bias=True) (fc2): Linear(in_features=128, out_features=10, bias=True) )
# nn.CrossEntropyLoss() はソフトマックス関数+クロスエントロピー誤差
criterion = nn.CrossEntropyLoss()
optimizer = optim.Adam(net.parameters(), lr=0.001)
# モデルを学習させる関数を作成
def train_model(net, dataloaders_dict, criterion, optimizer, num_epochs):
# epochのループ
for epoch in range(num_epochs):
print('Epoch {}/{}'.format(epoch+1, num_epochs))
print('-------------')
# epochごとの学習と検証のループ
for phase in ['train', 'val']:
if phase == 'train':
net.train() # モデルを訓練モードに
else:
net.eval() # モデルを検証モードに
epoch_loss = 0.0 # epochの損失和
epoch_corrects = 0 # epochの正解数
# 未学習時の検証性能を確かめるため、epoch=0の訓練は省略
if (epoch == 0) and (phase == 'train'):
continue
# データローダーからミニバッチを取り出すループ
for i , (inputs, labels) in tqdm(enumerate(dataloaders_dict[phase])):
# optimizerを初期化
optimizer.zero_grad()
# 順伝搬(forward)計算
with torch.set_grad_enabled(phase == 'train'): # 訓練モードのみ勾配を算出
outputs = net(inputs) # 順伝播
loss = criterion(outputs, labels) # 損失を計算
_, preds = torch.max(outputs, 1) # ラベルを予測
# 訓練時はバックプロパゲーション
if phase == 'train':
loss.backward()
optimizer.step()
# イタレーション結果の計算
# lossの合計を更新
epoch_loss += loss.item() * inputs.size(0)
# 正解数の合計を更新
epoch_corrects += torch.sum(preds == labels.data)
# epochごとのlossと正解率を表示
epoch_loss = epoch_loss / len(dataloaders_dict[phase].dataset)
epoch_acc = epoch_corrects.double() / len(dataloaders_dict[phase].dataset)
print('{} Loss: {:.4f} Acc: {:.4f}'.format(phase, epoch_loss, epoch_acc))
# 学習・検証を実行する
num_epochs = 3
train_model(net, dataloaders_dict, criterion, optimizer, num_epochs=num_epochs)
3it [00:00, 23.63it/s]
Epoch 1/3 -------------
157it [00:05, 30.87it/s] 1it [00:00, 9.87it/s]
val Loss: 2.3047 Acc: 0.1128 Epoch 2/3 -------------
938it [01:11, 13.09it/s] 3it [00:00, 28.23it/s]
train Loss: 0.1247 Acc: 0.9617
157it [00:04, 33.20it/s] 2it [00:00, 13.95it/s]
val Loss: 0.0486 Acc: 0.9844 Epoch 3/3 -------------
938it [01:26, 10.81it/s] 3it [00:00, 26.54it/s]
train Loss: 0.0447 Acc: 0.9862
157it [00:05, 26.97it/s]
val Loss: 0.0338 Acc: 0.9894
batch_iterator = iter(dataloaders_dict["val"]) # イテレータに変換
imges, labels = next(batch_iterator) # 1番目の要素を取り出す
net.eval() #推論モード
with torch.set_grad_enabled(False): # 推論モードでは勾配を算出しない
outputs = net(imges) # 順伝播
_, preds = torch.max(outputs, 1) # ラベルを予測
#テストデータの予測結果を描画
plt.imshow(imges[0].numpy().reshape(28,28), cmap='gray')
plt.title("Label: Target={}, Predict={}".format(labels[0], preds[0].numpy()))
plt.show()
以上