Target 40 is out of bounds for nn.CrossEntropyLoss() - pandas

Question

I create a custom image data set like:
from torch.utils.data.dataset import Dataset
from PIL import Image
import torchvision
from torchvision import datasets, models, transforms
import numpy as np
class MyCustomDataset(Dataset):
def __init__(self, df, transforms=None):
"""
Args:
df (pandas.DataFrame): path to images and in dataframe
transforms: pytorch transforms for transforms and tensor conversion
"""
# Transforms
self.transforms = transforms
# Read the dataframe
self.data_info = df
# First column contains the image paths
self.image_arr = np.asarray(self.data_info.iloc[:, 0])
# Second column is the labels
self.label_arr = np.asarray(self.data_info.iloc[:, 1])
# Calculate len
self.data_len = len(self.data_info.index)
def __getitem__(self, index):
# Get image name from the pandas df
single_image_name = self.image_arr[index]
# Open image
img_as_img = Image.open(single_image_name)
img_as_tensor= self.transforms(img_as_img)
# Get label(class) of the image based on the cropped pandas column
single_image_label = self.label_arr[index]
return (img_as_tensor, single_image_label)
def __len__(self):
return self.data_len
The input of df for MyCustomDataset(df, transforms) is pandas dataframe storing images' paths and labels look like below:
file_name label
0 M:\RealModels\images\001\001001.png 0
1 M:\RealModels\images\001\002001.png 0
2 M:\RealModels\images\001\003001.png 0
3 M:\RealModels\images\001\004001.png 0
4 M:\RealModels\images\001\006001.png 0
... ... ...
3197 M:\RenderedModels\images_rgb\450\116450.png 45
3198 M:\RenderedModels\images_rgb\450\117450.png 45
3199 M:\RenderedModels\images_rgb\450\118450.png 45
3200 M:\RenderedModels\images_rgb\450\119450.png 45
3201 M:\RenderedModels\images_rgb\450\120450.png 45
3202 rows × 2 columns
There are 16 classes in my dataset. The classes lables are like:['00', '01', '12', '34','35'...,'45']
My whole program is:
import torch
import torch.nn as nn
import torch.optim as optim
from torch.optim import lr_scheduler
import numpy as np
from PIL import Image
import matplotlib.pyplot as plt
import time
import os
import copy
import scipy
from torch.utils.tensorboard import SummaryWriter
from torch.utils.data.dataset import Dataset
import torchvision
from torchvision import datasets, models, transforms
# In[2]:
import sys
sys.path.append(r"M:\program\pytorch\Scripts")
import custom_fun
from custom_fun import custom_dataset
import create_folder
from create_folder import create_tb_folder
import dataset_from_image
from dataset_from_image import MyCustomDataset
# In[3]:
tb_dir = r'../'
path = create_tb_folder(tb_dir)
# In[4]:
mean = np.array([0.485, 0.456, 0.406])
std = np.array([0.229, 0.224, 0.225])
# In[5]:
data_transforms = {
'train': transforms.Compose([
transforms.RandomRotation(degrees=15),
transforms.Resize((224,224)), # 299 for Inception v3
transforms.ColorJitter(),
transforms.ToTensor(),
transforms.Normalize(mean, std)
]),
'val': transforms.Compose([
transforms.Resize((224,224)),
transforms.ToTensor(),
transforms.Normalize(mean, std)
]),
}
# In[6]:
data_dir = r'M:\dataset\first_att'
# In[7]:
batch_size = 4
lr = 0.003
# In[8]:
data = custom_dataset(data_dir,0.8)
# In[37]:
df = data['train']
df.head()
# In[9]:
image_datasets = {x: MyCustomDataset(data[x], data_transforms[x]) for x in ['train', 'val']}
# In[24]:
dataloaders = {x: torch.utils.data.DataLoader(image_datasets[x], batch_size=batch_size,
shuffle=True, num_workers=0)
for x in ['train', 'val']}
# In[25]:
dataset_sizes = {x: len(image_datasets[x]) for x in ['train', 'val']}
device = torch.device("cpu")
# In[26]:
# Get a batch of training data
inputs, classes = next(iter(dataloaders['train']))
print(inputs.shape,classes.shape)
print(classes)
# Make a grid from batch
out = torchvision.utils.make_grid(inputs)
# In[27]:
tb1 = SummaryWriter(path[1])
tb2 = SummaryWriter(path[2])
# In[28]:
def train_model(model, criterion, optimizer, scheduler, num_epochs):
since = time.time()
best_model_wts = copy.deepcopy(model.state_dict())
best_acc = 0.0
for epoch in range(num_epochs):
print('Epoch {}/{}'.format(epoch, num_epochs - 1))
print('-' * 10)
# Each epoch has a training and validation phase
for phase in ['train', 'val']:
if phase == 'train':
model.train() # Set model to training mode
else:
model.eval() # Set model to evaluate mode
running_loss = 0.0
running_corrects = 0
# Iterate over data.
for inputs, labels in dataloaders[phase]:
inputs = inputs.to(device)
labels = labels.to(device)
# forward
# track history if only in train
with torch.set_grad_enabled(phase == 'train'):
outputs = model(inputs)
_, preds = torch.max(outputs, 1)
loss = criterion(outputs, labels)
# backward + optimize only if in training phase
if phase == 'train':
optimizer.zero_grad()
loss.backward()
optimizer.step()
# statistics
running_loss += loss.item() * inputs.size(0)
running_corrects += torch.sum(preds == labels.data)
if phase == 'train':
scheduler.step()
epoch_loss = running_loss / dataset_sizes[phase] # images of training data
epoch_acc = running_corrects.double() / dataset_sizes[phase] # images of val data
print('{} Loss: {:.4f} Acc: {:.4f}'.format(
phase, epoch_loss, epoch_acc))
# Tensorboard works here
if phase == 'train':
tb1.add_scalar('Loss', epoch_loss, epoch)
tb1.add_scalar('Accuracy', epoch_acc, epoch)
for name, weight in model.named_parameters():
tb.add_histogram(name, weight, epoch)
tb.add_histogram(f'{name}.grad', weight.grad, epoch)
else:
tb2.add_scalar('Loss', epoch_loss, epoch)
tb2.add_scalar('Accuracy', epoch_acc, epoch)
# deep copy the model
if phase == 'val' and epoch_acc > best_acc:
best_acc = epoch_acc
best_model_wts = copy.deepcopy(model.state_dict())
time_elapsed = time.time() - since
print('Training complete in {:.0f}m {:.0f}s'.format(
time_elapsed // 60, time_elapsed % 60))
print('Best val Acc: {:4f}'.format(best_acc))
# load best model weights
model.load_state_dict(best_model_wts)
return model
# In[29]:
model = models.resnet18(pretrained=True)
# The way below is feature extraction.
#for param in model.parameters():
#param.requires_grad = False
# In[30]:
num_ftrs = model.fc.in_features
# In[31]:
model.fc = nn.Linear(num_ftrs, 16)
# In[32]:
model = model.to(device)
criterion = nn.CrossEntropyLoss()
# In[33]:
# Observe that all parameters are being optimized
optimizer = optim.Adam(model.parameters(), lr=lr)
# In[34]:
step_lr_scheduler = lr_scheduler.StepLR(optimizer, step_size=5, gamma=0.1)
# In[35]:
tb = SummaryWriter(path[0])
grid = torchvision.utils.make_grid(inputs)
tb.add_image('images', grid)
tb.add_graph(model, inputs)
tb.close()
# In[36]:
model = train_model(model, criterion, optimizer, step_lr_scheduler, num_epochs=20)
# In[ ]:
torch.save(model.state_dict(), 'first_att_02.pth')
# In[ ]:
But I got the error:
---------------------------------------------------------------------------
IndexError Traceback (most recent call last)
<ipython-input-36-edba91d1cb93> in <module>
----> 1 model = train_model(model, criterion, optimizer, step_lr_scheduler, num_epochs=20)
<ipython-input-28-662f652902cf> in train_model(model, criterion, optimizer, scheduler, num_epochs)
29 outputs = model(inputs)
30 _, preds = torch.max(outputs, 1)
---> 31 loss = criterion(outputs, labels)
32
33 # backward + optimize only if in training phase
M:\program\pytorch\lib\site-packages\torch\nn\modules\module.py in __call__(self, *input, **kwargs)
530 result = self._slow_forward(*input, **kwargs)
531 else:
--> 532 result = self.forward(*input, **kwargs)
533 for hook in self._forward_hooks.values():
534 hook_result = hook(self, input, result)
M:\program\pytorch\lib\site-packages\torch\nn\modules\loss.py in forward(self, input, target)
914 def forward(self, input, target):
915 return F.cross_entropy(input, target, weight=self.weight,
--> 916 ignore_index=self.ignore_index, reduction=self.reduction)
917
918
M:\program\pytorch\lib\site-packages\torch\nn\functional.py in cross_entropy(input, target, weight, size_average, ignore_index, reduce, reduction)
2019 if size_average is not None or reduce is not None:
2020 reduction = _Reduction.legacy_get_string(size_average, reduce)
-> 2021 return nll_loss(log_softmax(input, 1), target, weight, None, ignore_index, None, reduction)
2022
2023
M:\program\pytorch\lib\site-packages\torch\nn\functional.py in nll_loss(input, target, weight, size_average, ignore_index, reduce, reduction)
1836 .format(input.size(0), target.size(0)))
1837 if dim == 2:
-> 1838 ret = torch._C._nn.nll_loss(input, target, weight, _Reduction.get_enum(reduction), ignore_index)
1839 elif dim == 4:
1840 ret = torch._C._nn.nll_loss2d(input, target, weight, _Reduction.get_enum(reduction), ignore_index)
IndexError: Target 40 is out of bounds.
Could you let me know where I am wrong? Thanks very much.