【貓狗數據集】加載保存的模型進行測試

已重新上傳好數據集：

分割線-----------------------------------------------------------------

2020.3.10

發現數據集沒有完整的上傳到谷歌的colab上去，我說怎么計算出來的step不對勁。

測試集是完整的。

訓練集中cat的確是有10125張圖片，而dog只有1973張，所以完成一個epoch需要迭代的次數為：

(10125+1973)/128=94.515625，約等於95。

順便提一下，有兩種方式可以計算出數據集的量：

第一種：print(len(train_dataset))

第二種：在../dog目錄下，輸入ls | wc -c

今天重新上傳dog數據集。

分割線-----------------------------------------------------------------

數據集下載地址：

鏈接：https://pan.baidu.com/s/1l1AnBgkAAEhh0vI5_loWKw
提取碼：2xq4

創建數據集：https://www.cnblogs.com/xiximayou/p/12398285.html

讀取數據集：https://www.cnblogs.com/xiximayou/p/12422827.html

進行訓練：https://www.cnblogs.com/xiximayou/p/12448300.html

保存模型並繼續進行訓練：https://www.cnblogs.com/xiximayou/p/12452624.html

epoch、batchsize、step之間的關系：https://www.cnblogs.com/xiximayou/p/12405485.html

 

我們在test目錄下新建一個文件test.py

test.py

import sys
sys.path.append("/content/drive/My Drive/colab notebooks")
from utils import rdata
from model import resnet
import torch.nn as nn
import torch
import numpy as np
import torchvision


device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')

train_loader,test_loader,train_data,test_data=rdata.load_dataset()
model =torchvision.models.resnet18(pretrained=False)
model.fc = nn.Linear(model.fc.in_features,2,bias=False)
model.cuda()
#print(model) 

save_path="/content/drive/My Drive/colab notebooks/output/dogcat-resnet18.t7" 
checkpoint = torch.load(save_path)
model.load_state_dict(checkpoint['model'])
start_epoch = checkpoint['epoch']
start_loss=checkpoint["train_loss"]
start_acc=checkpoint["train_acc"]
print("當前epoch:{} 當前訓練損失：{:.4f} 當前訓練准確率：{:.4f}".format(start_epoch+1,start_loss,start_acc))

num_epochs=1
criterion=nn.CrossEntropyLoss()

# Train the model
total_step = len(test_loader)
def test():
  for epoch in range(num_epochs):
      tot_loss = 0.0
      correct = 0
      for i ,(images, labels) in enumerate(test_loader):
          images = images.cuda()
          labels = labels.cuda()

          # Forward pass
          outputs = model(images)
          _, preds = torch.max(outputs.data,1)
          loss = criterion(outputs, labels)
          tot_loss += loss.data
          correct += torch.sum(preds == labels.data).to(torch.float32)
          if (i+1) % 2 == 0:
              print('Epoch: [{}/{}], Step: [{}/{}], Loss: {:.4f}'
                    .format(epoch+1, num_epochs, i+1, total_step, loss.item()))
      ### Epoch info ####
      epoch_loss = tot_loss/len(test_data)
      print('test loss: {:.4f}'.format(epoch_loss))
      epoch_acc = correct/len(test_data)
      print('test acc: {:.4f}'.format(epoch_acc))
with torch.no_grad(): test()

需要注意，測試的時候我們不需要進行反向傳播更新參數。

結果：

當前epoch:2 當前訓練損失：0.0037 當前訓練准確率：0.8349
Epoch: [1/1], Step: [2/38], Loss: 1.0218
Epoch: [1/1], Step: [4/38], Loss: 0.9890
Epoch: [1/1], Step: [6/38], Loss: 0.9255
Epoch: [1/1], Step: [8/38], Loss: 0.9305
Epoch: [1/1], Step: [10/38], Loss: 0.9013
Epoch: [1/1], Step: [12/38], Loss: 1.0436
Epoch: [1/1], Step: [14/38], Loss: 0.8102
Epoch: [1/1], Step: [16/38], Loss: 0.9356
Epoch: [1/1], Step: [18/38], Loss: 0.8668
Epoch: [1/1], Step: [20/38], Loss: 1.0083
Epoch: [1/1], Step: [22/38], Loss: 1.0202
Epoch: [1/1], Step: [24/38], Loss: 0.8906
Epoch: [1/1], Step: [26/38], Loss: 1.0110
Epoch: [1/1], Step: [28/38], Loss: 0.8508
Epoch: [1/1], Step: [30/38], Loss: 0.9539
Epoch: [1/1], Step: [32/38], Loss: 0.9225
Epoch: [1/1], Step: [34/38], Loss: 0.9501
Epoch: [1/1], Step: [36/38], Loss: 0.8252
Epoch: [1/1], Step: [38/38], Loss: 0.9201
test loss: 0.0074
test acc: 0.5000