NameError Traceback (most recent call last)
in
21
22 # 학습용 이미지를 무작위로 가져오기
—> 23 dataiter = iter(trainloader)
24 images, labels = dataiter.next()
25
NameError: name ‘trainloader’ is not defined
This is a full error message. Pleae help.
This error is raised, if the variable trainloader is undefined. Often this could happen, if you have a typo in the name or the code logic is wrong, e.g. the trainloader is defined after its first usage.
Thanks for your answer.
but as you can see, ‘trainloader’ is defiend…
Also, I was running the example, not the code I wrote myself.
I don’t know what might be wrong with this code.
Could you post an executable code snippet to reproduce this issue?
[1]
import torch
import torchvision
import torchvision.transforms as transforms
import torch.utils.data as data
import torchvision.datasets as datasets
import torch.nn as nn
[2]
transform = transforms.Compose(
[transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))])
[3]
train_data = datasets.CIFAR10(root = “./data”,
train = True,
download = True,
transform = transform)
train_data, val_data = torch.utils.data.random_split(train_data, [int(len(train_data) * 0.8), int(len(train_data)*0.2)])
test_data = datasets.CIFAR10(root = “./data”,
train = False,
download = True,
transform = transform)
[4]
classes = test_data.classes
dic_classes = {}
for i in range(len(classes)):
dic_classes[i] = classes[i]
print(dic_classes)
[5]
trainloader = torch.utils.data.DataLoader(train_data, batch_size=16,
shuffle=True)
valloader = torch.utils.data.DataLoader(val_data, batch_size=16,
shuffle=True)
testloader = torch.utils.data.DataLoader(test_data, batch_size=16,
shuffle=False)
[6]
import matplotlib.pyplot as plt
import numpy as np
def imshow(img, labels, dic):
num = len(labels)
rows = int(np.sqrt(num))
cols = int(np.sqrt(num))
fig = plt.figure(figsize=(20,20))
for i in range(rows*cols):
ax = fig.add_subplot(rows, cols, i+1)
tmp = img[i]
tmp = tmp / 2 + 0.5 # unnormalize
npimg = tmp.numpy()
ax.imshow(np.transpose(npimg, (1, 2, 0)), cmap = "bone")
ax.title.set_text(dic[labels[i].item()])
#plt.show()
ax.axis('off')
dataiter = iter(trainloader)
images, labels = dataiter.next()
imshow(images, labels, dic_classes)
‘[6]’ is issued code. I hope this helps.
[1]~[5] all worked normally. The mentioned error occurs in [6].
Thanks for the code.
Your code works fine on my machine. Here is the formatted code in case you want to rerun it:
import torch
import torchvision
import torchvision.transforms as transforms
import torch.utils.data as data
import torchvision.datasets as datasets
import torch.nn as nn
transform = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
])
train_data = datasets.CIFAR10(
root = "./data",
train = True,
download = True,
transform = transform)
train_data, val_data = torch.utils.data.random_split(
train_data, [int(len(train_data) * 0.8), int(len(train_data)*0.2)])
test_data = datasets.CIFAR10(
root = "./data",
train = False,
download = True,
transform = transform)
classes = test_data.classes
dic_classes = {}
for i in range(len(classes)):
dic_classes[i] = classes[i]
print(dic_classes)
trainloader = torch.utils.data.DataLoader(
train_data, batch_size=16, shuffle=True)
valloader = torch.utils.data.DataLoader(
val_data, batch_size=16, shuffle=True)
testloader = torch.utils.data.DataLoader(
test_data, batch_size=16, shuffle=False)
import matplotlib.pyplot as plt
import numpy as np
def imshow(img, labels, dic):
num = len(labels)
rows = int(np.sqrt(num))
cols = int(np.sqrt(num))
fig = plt.figure(figsize=(20,20))
for i in range(rows*cols):
ax = fig.add_subplot(rows, cols, i+1)
tmp = img[i]
tmp = tmp / 2 + 0.5 # unnormalize
npimg = tmp.numpy()
ax.imshow(np.transpose(npimg, (1, 2, 0)), cmap = "bone")
ax.title.set_text(dic[labels[i].item()])
#plt.show()
ax.axis('off')
dataiter = iter(trainloader)
images, labels = dataiter.next()
imshow(images, labels, dic_classes)
I don’t get any errors and the sample images are displayed.
PS: you can post code snippets by warpping them into three backticks ```, which makes debugging easier. 
for images, labels in train_loader:
pass
images, labels = next(iter(train_loader))
indx=10
plt.imshow(images[indx].reshape(64,64))
plt.title(label_map[int(labels[indx].numpy())])
Hello, I hope you are fine. I am facing the same issue “NameError: name ‘train_loader’ is not defined”, Here’s my code. Can you help me to resolve it. I will be very thankful to you.
"import argparse
import os
import pickle
import time
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import PIL
import torch
import torch.optim as optim
import torchvision
from PIL import Image
from torch import nn
from torch.nn import functional as F
from datasets import QProper_signals1, QProper_signals2, QImproper_signals1, QImproper_signals2
from models.vanilla_vae_q import QuaternionFCVAE2
parser = argparse.ArgumentParser()
parser.add_argument(“–gpu_num”, type=int, default=0)
parser.add_argument(“–track_signals”, type=bool, default=False)
parser.add_argument(
“–experiment”, type=int, default=3, help=“1, 2”
)
parser.add_argument(“–len_train”, type=int, default=1000)
parser.add_argument(“–len_val”, type=int, default=500)
parser.add_argument(“–n”, type=int, default=1)
parser.add_argument(
“–signal”, type=str, default=“improper”, help=“Either proper or improper.”
)
parser.add_argument(
“–n_channels”,
type=int,
default=4,
help=“3 for real-valued inputs, 4 for quaternion-valued inputs.”,
)
parser.add_argument(
“–latent_dim”, type=int, default=100, help=“Dimension of the latent space.”
)
parser.add_argument(“–n_epochs”, type=int, default=5000, help=“Training epochs.”)
parser.add_argument(“–batch_size”, type=int, default=1, help=“Batch size.”)
parser.add_argument(“–lr”, type=float, default=0.0005, help=“Learning rate.”)
parser.add_argument(“–kld_weight”, type=float, default=1, help=“Weight KLD in loss.”)
parser.add_argument(“–cuda”, type=bool, default=True)
parser.add_argument(
“–num_samples”, type=int, default=1000, help=“Number of samples to generate.”
)
opt = parser.parse_args()
print(“#######”)
print(
“Running experiment %i on %s signals for %i epochs, with KLD weight %f”
% (opt.experiment, opt.signal, opt.n_epochs, opt.kld_weight)
)
print(“######”)
def loss_function(recons, input, mu, log_var, kld_weight=opt.kld_weight):
“”“Computes QVAE loss function.
QVAEloss function is composed of BCE reconstruction loss and weighted KL divergence.
Quaternion-valued KL divergence is defined according to the paper.”“”
# recons_loss = F.binary_cross_entropy(recons, input)
recons_loss = F.mse_loss(recons, input)
kld_loss = torch.mean(
0.5 * (torch.sum(log_var.exp() + mu ** 2 - 1, dim=2))
- 2 * torch.sum(log_var, dim=2),
dim=0,
)
kld_loss = torch.mean(kld_loss)
loss = recons_loss + opt.kld_weight * kld_loss
return loss, recons_loss, -kld_loss
###############################
###############################
qvae = QuaternionFCVAE2(
in_channels=opt.n_channels, latent_dim=opt.latent_dim, n=opt.n
)
if opt.experiment == 1:
# Define input signal
if opt.signal == “proper”:
train_loader, val_loader, _ = QProper_signals1(
opt.len_train, opt.len_val, 100, opt.n
)
elif opt.signal == “improper”:
train_loader, val_loader, _ = QImproper_signals1(
opt.len_train, opt.len_val, 100, opt.n
)
else:
print(“Wrong kind of signal. Signal can be either ‘proper’ or ‘improper’.”)
elif opt.experiment == 2:
# Define input signal
if opt.signal == “proper”:
train_loader, val_loader, _ = QProper_signals2(
opt.len_train, opt.len_val, 100, opt.n
)
elif opt.signal == “improper”:
train_loader, val_loader, _ = QImproper_signals2(
opt.len_train, opt.len_val, 100, opt.n
)
else:
print(“Wrong kind of signal. Signal can be either ‘proper’ or ‘improper’.”)
print(
“Number of parameters”, sum(p.numel() for p in qvae.parameters() if p.requires_grad)
)
device = torch.device(“cuda” if torch.cuda.is_available() and opt.cuda else “cpu”)
qvae.to(device)
optimizer = optim.Adam(qvae.parameters(), lr=opt.lr)
loss_train =
loss_val =
kld_train =
kld_val =
device = torch.device(“cpu”)
if opt.cuda:
device = “cuda:%i” % opt.gpu_num
else:
device = “cpu”
###########################
###########################
def train(model):
model.train()
start = time.time()
for epoch in range(opt.n_epochs):
# Training
temp_loss_train =
temp_kld_train =
for idx, data in enumerate(train_loader):
data = data.to(device)
optimizer.zero_grad()
# Apply model
recons, input, mu, log_var = model(data)
# Compute loss function
loss, _, kld = loss_function(recons, input, mu, log_var)
kld_train.append(kld.detach().item())
temp_kld_train.append(kld.detach().item())
loss_train.append(loss.detach().item())
temp_loss_train.append(loss.detach().item())
loss.backward()
optimizer.step()
# Validation
with torch.no_grad():
temp_loss_val = []
temp_kld_val = []
for val_data in val_loader:
val_data = val_data.to(device)
# Evaluate model
val_recons, val_input, val_mu, val_log_var = model(val_data)
# Compute loss function
loss, _, kld = loss_function(val_recons, val_input, val_mu, val_log_var)
kld_val.append(kld.detach().item())
temp_kld_val.append(kld.detach().item())
temp_loss_val.append(loss.item())
loss_val.append(loss.item())
if epoch % 10 == 0:
if opt.track_signals:
gen_samples = model.sample(
num_samples=opt.num_samples, current_device=opt.gpu_num
)
df = pd.DataFrame(gen_samples.detach().cpu().numpy())
df.to_csv(
"./signals/toy_example%i/track_signals/%s_samples_%i.csv"
% (opt.experiment, opt.signal, epoch),
index=False, header=None
)
end = time.time()
print(
"[Epoch: %i][Train loss: %f][Train KLD: %f][Val loss: %f][Val KLD: %f][Time: %f]"
% (
epoch,
np.mean(temp_loss_train),
np.mean(temp_kld_train),
np.mean(temp_loss_val),
np.mean(temp_kld_val),
end - start,
)
)
# scheduler.step()
# # Store losses
# torch.save(loss_train, 'checkpoints/toy_example2/loss_train_qproper')
# torch.save(loss_val, 'checkpoints/toy_example2/loss_val_qproper')
# torch.save(kld_train, 'checkpoints/toy_example2/kld_train_qproper')
# torch.save(kld_val, 'checkpoints/toy_example2/kld_val_qproper')
# Store model
torch.save(
model.state_dict(),
"./%i/model_%s_epoch%i" % (opt.experiment, opt.signal, epoch),
)
train(qvae)
"
I guess the conditions checking the signal to create the train_loader might not be met and this object is thus never created. You could error out instead of printing a message only.
Since your code is not properly formatted I’m unable to run and debug it further.