Predictive modeling with deep learning is a skill that modern developers need to know.
PyTorch is the premier open-source deep learning framework developed and maintained by Facebook.
At its core, PyTorch is a mathematical library that allows you to perform efficient computation and automatic differentiation on graph-based models. Achieving this directly is challenging, although thankfully, the modern PyTorch API provides classes and idioms that allow you to easily develop a suite of deep learning models.
In this tutorial, you will discover a step-by-step guide to developing deep learning models in PyTorch.
After completing this tutorial, you will know:
The difference between Torch and PyTorch and how to install and confirm PyTorch is working.
The five-step life-cycle of PyTorch models and how to define, fit, and evaluate models.
How to develop PyTorch deep learning models for regression, classification, and predictive modeling tasks.
Kick-start your project with my bookDeep Learning with PyTorch. It providesself-study tutorials withworking code.
Let’s get started.
PyTorch Tutorial – How to Develop Deep Learning Models Photo byDimitry B., some rights reserved.
PyTorch Tutorial Overview
The focus of this tutorial is on using the PyTorch API for common deep learning model development tasks; we will not be diving into the math and theory of deep learning. For that, I recommendstarting with this excellent book.
The best way to learn deep learning in python is by doing. Dive in. You can circle back for more theory later.
I have designed each code example to use best practices and to be standalone so that you can copy and paste it directly into your project and adapt it to your specific needs. This will give you a massive head start over trying to figure out the API from official documentation alone.
It is a large tutorial, and as such, it is divided into three parts; they are:
How to Install PyTorch
What Are Torch and PyTorch?
How to Install PyTorch
How to Confirm PyTorch Is Installed
PyTorch Deep Learning Model Life-Cycle
Step 1: Prepare the Data
Step 2: Define the Model
Step 3: Train the Model
Step 4: Evaluate the Model
Step 5: Make Predictions
How to Develop PyTorch Deep Learning Models
How to Develop an MLP for Binary Classification
How to Develop an MLP for Multiclass Classification
How to Develop an MLP for Regression
How to Develop a CNN for Image Classification
You Can Do Deep Learning in Python!
Work through this tutorial. It will take you 60 minutes, max!
You do not need to understand everything (at least not right now). Your goal is to run through the tutorial end-to-end and get a result. You do not need to understand everything on the first pass. List down your questions as you go. Make heavy use of the API documentation to learn about all of the functions that you’re using.
You do not need to know the math first. Math is a compact way of describing how algorithms work, specifically tools from linear algebra, probability, and calculus. These are not the only tools that you can use to learn how algorithms work. You can also use code and explore algorithm behavior with different inputs and outputs. Knowing the math will not tell you what algorithm to choose or how to best configure it. You can only discover that through carefully controlled experiments.
You do not need to know how the algorithms work. It is important to know about the limitations and how to configure deep learning algorithms. But learning about algorithms can come later. You need to build up this algorithm knowledge slowly over a long period of time. Today, start by getting comfortable with the platform.
You do not need to be a Python programmer. The syntax of the Python language can be intuitive if you are new to it. Just like other languages, focus on function calls (e.g. function()) and assignments (e.g. a = “b”). This will get you most of the way. You are a developer; you know how to pick up the basics of a language really fast. Just get started and dive into the details later.
You do not need to be a deep learning expert. You can learn about the benefits and limitations of various algorithms later, and there are plenty of tutorials that you can read to brush up on the steps of a deep learning project.
1. How to Install PyTorch
In this section, you will discover what PyTorch is, how to install it, and how to confirm that it is installed correctly.
1.1. What Are Torch and PyTorch?
PyTorch is an open-source Python library for deep learning developed and maintained by Facebook.
The project started in 2016 and quickly became a popular framework among developers and researchers.
Torch (Torch7) is an open-source project for deep learning written in C and generally used via the Lua interface. It was a precursor project to PyTorch and is no longer actively developed. PyTorch includes “Torch” in the name, acknowledging the prior torch library with the “Py” prefix indicating the Python focus of the new project.
The PyTorch API is simple and flexible, making it a favorite for academics and researchers in the development of new deep learning models and applications. The extensive use has led to many extensions for specific applications (such as text, computer vision, and audio data), and may pre-trained models that can be used directly. As such, it may be the most popular library used by academics.
The flexibility of PyTorch comes at the cost of ease of use, especially for beginners, as compared to simpler interfaces likeKeras. The choice to use PyTorch instead of Keras gives up some ease of use, a slightly steeper learning curve, and more code for more flexibility, and perhaps a more vibrant academic community.
1.2. How to Install PyTorch
Before installing PyTorch, ensure that you have Python installed, such as Python 3.6 or higher.
If you don’t have Python installed, you can install it using Anaconda. This tutorial will show you how:
There are many ways to install the PyTorch open-source deep learning library.
The most common, and perhaps simplest, way to install PyTorch on your workstation is by using pip.
For example, on the command line, you can type:
1
sudo pip install torch
Perhaps the most popular application of deep learning is forcomputer vision, and the PyTorch computer vision package is called “torchvision.”
Installing torchvision is also highly recommended and it can be installed as follows:
1
sudo pip install torchvision
If you prefer to use an installation method more specific to your platform or package manager, you can see a complete list of installation instructions here:
All examples in this tutorial will work just fine on a modern CPU. If you want to configure PyTorch for your GPU, you can do that after completing this tutorial. Don’t get distracted!
1.3. How to Confirm PyTorch Is Installed
Once PyTorch is installed, it is important to confirm that the library was installed successfully and that you can start using it.
Don’t skip this step.
If PyTorch is not installed correctly or raises an error on this step, you won’t be able to run the examples later.
Create a new file calledversions.py and copy and paste the following code into the file.
1
2
3
# check pytorch version
importtorch
print(torch.__version__)
Save the file, then open your command line and change directory to where you saved the file.
Then type:
1
python versions.py
You should then see output like the following:
1
1.3.1
This confirms that PyTorch is installed correctly and that we are all using the same version.
This also shows you how to run a Python script from the command line. I recommend running all code from the command line in this manner, and not from a notebook or an IDE.
2. PyTorch Deep Learning Model Life-Cycle
In this section, you will discover the life-cycle for a deep learning model and the PyTorch API that you can use to define models.
A model has a life-cycle, and this very simple knowledge provides the backbone for both modeling a dataset and understanding the PyTorch API.
The five steps in the life-cycle are as follows:
1. Prepare the Data.
2. Define the Model.
3. Train the Model.
4. Evaluate the Model.
5. Make Predictions.
Let’s take a closer look at each step in turn.
Note: There are many ways to achieve each of these steps using the PyTorch API, although I have aimed to show you the simplest, or most common, or most idiomatic.
If you discover a better approach, let me know in the comments below.
Step 1: Prepare the Data
The first step is to load and prepare your data.
Neural network models require numerical input data and numerical output data.
You can use standard Python libraries to load and prepare tabular data, like CSV files. For example, Pandas can be used to load your CSV file, and tools from scikit-learn can be used to encode categorical data, such as class labels.
PyTorch provides theDataset class that you can extend and customize to load your dataset.
For example, the constructor of your dataset object can load your data file (e.g. a CSV file). You can then override the__len__() function that can be used to get the length of the dataset (number of rows or samples), and the__getitem__() function that is used to get a specific sample by index.
When loading your dataset, you can also perform any required transforms, such as scaling or encoding.
A skeleton of a customDataset class is provided below.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
# dataset definition
classCSVDataset(Dataset):
# load the dataset
def__init__(self,path):
# store the inputs and outputs
self.X=...
self.y=...
# number of rows in the dataset
def__len__(self):
returnlen(self.X)
# get a row at an index
def__getitem__(self,idx):
return[self.X[idx],self.y[idx]]
Once loaded, PyTorch provides theDataLoader class to navigate aDataset instance during the training and evaluation of your model.
ADataLoader instance can be created for the training dataset, test dataset, and even a validation dataset.
Therandom_split() function can be used to split a dataset into train and test sets. Once split, a selection of rows from theDataset can be provided to a DataLoader, along with the batch size and whether the data should be shuffled every epoch.
For example, we can define aDataLoader by passing in a selected sample of rows in the dataset.
Once defined, aDataLoader can be enumerated, yielding one batch worth of samples each iteration.
1
2
3
4
...
# train the model
fori,(inputs,targets)inenumerate(train_dl):
...
Step 2: Define the Model
The next step is to define a model.
The idiom for defining a model in PyTorch involves defining a class that extends theModule class.
The constructor of your class defines the layers of the model and the forward() function is the override that defines how to forward propagate input through the defined layers of the model.
Many layers are available, such asLinear for fully connected layers,Conv2d for convolutional layers, andMaxPool2d for pooling layers.
Activation functions can also be defined as layers, such asReLU,Softmax, andSigmoid.
Below is an example of a simple MLP model with one layer.
1
2
3
4
5
6
7
8
9
10
11
12
13
# model definition
classMLP(Module):
# define model elements
def__init__(self,n_inputs):
super(MLP,self).__init__()
self.layer=Linear(n_inputs,1)
self.activation=Sigmoid()
# forward propagate input
defforward(self,X):
X=self.layer(X)
X=self.activation(X)
returnX
The weights of a given layer can also be initialized after the layer is defined in the constructor.
Common examples include theXavier andHe weight initialization schemes. For example:
1
2
...
xavier_uniform_(self.layer.weight)
Step 3: Train the Model
The training process requires that you define a loss function and an optimization algorithm.
Common loss functions include the following:
BCELoss: Binary cross-entropy loss for binary classification.
CrossEntropyLoss: Categorical cross-entropy loss for multi-class classification.
Stochastic gradient descent is used for optimization, and the standard algorithm is provided by theSGD class, although other versions of the algorithm are available, such asAdam.
Training the model involves enumerating theDataLoader for the training dataset.
First, a loop is required for the number of training epochs. Then an inner loop is required for the mini-batches for stochastic gradient descent.
1
2
3
4
5
6
...
# enumerate epochs
forepochinrange(100):
# enumerate mini batches
fori,(inputs,targets)inenumerate(train_dl):
...
Each update to the model involves the same general pattern comprised of:
Clearing the last error gradient.
A forward pass of the input through the model.
Calculating the loss for the model output.
Backpropagating the error through the model.
Update the model in an effort to reduce loss.
For example:
1
2
3
4
5
6
7
8
9
10
11
...
# clear the gradients
optimizer.zero_grad()
# compute the model output
yhat=model(inputs)
# calculate loss
loss=criterion(yhat,targets)
# credit assignment
loss.backward()
# update model weights
optimizer.step()
Step 4: Evaluate the model
Once the model is fit, it can be evaluated on the test dataset.
This can be achieved by using theDataLoader for the test dataset and collecting the predictions for the test set, then comparing the predictions to the expected values of the test set and calculating a performance metric.
1
2
3
4
5
...
fori,(inputs,targets)inenumerate(test_dl):
# evaluate the model on the test set
yhat=model(inputs)
...
Step 5: Make predictions
A fit model can be used to make a prediction on new data.
For example, you might have a single image or a single row of data and want to make a prediction.
This requires that you wrap the data in aPyTorch Tensor data structure.
A Tensor is just the PyTorch version of a NumPy array for holding data. It also allows you to perform the automatic differentiation tasks in the model graph, like callingbackward() when training the model.
The prediction too will be a Tensor, although you can retrieve the NumPy array bydetaching the Tensor from the automatic differentiation graph and calling the NumPy function.
1
2
3
4
5
6
7
...
# convert row to data
row=Variable(Tensor([row]).float())
# make prediction
yhat=model(row)
# retrieve numpy array
yhat=yhat.detach().numpy()
Now that we are familiar with the PyTorch API at a high-level and the model life-cycle, let’s look at how we can develop some standard deep learning models from scratch.
3. How to Develop PyTorch Deep Learning Models
In this section, you will discover how to develop, evaluate, and make predictions with standard deep learning models, including Multilayer Perceptrons (MLP) and Convolutional Neural Networks (CNN).
A Multilayer Perceptron model, or MLP for short, is a standard fully connected neural network model.
It is comprised of layers of nodes where each node is connected to all outputs from the previous layer and the output of each node is connected to all inputs for nodes in the next layer.
An MLP is a model with one or more fully connected layers. This model is appropriate for tabular data, that is data as it looks in a table or spreadsheet with one column for each variable and one row for each variable. There are three predictive modeling problems you may want to explore with an MLP; they are binary classification, multiclass classification, and regression.
Let’s fit a model on a real dataset for each of these cases.
Note: The models in this section are effective, but not optimized. See if you can improve their performance. Post your findings in the comments below.
3.1. How to Develop an MLP for Binary Classification
We will use the Ionosphere binary (two class) classification dataset to demonstrate an MLP for binary classification.
This dataset involves predicting whether there is a structure in the atmosphere or not given radar returns.
The dataset will be downloaded automatically using Pandas, but you can learn more about it here.
We will use aLabelEncoder to encode the string labels to integer values 0 and 1. The model will be fit on 67 percent of the data, and the remaining 33 percent will be used for evaluation, split using thetrain_test_split() function.
It is a good practice to use ‘relu‘ activation with a ‘He Uniform‘ weight initialization. This combination goes a long way to overcome the problem ofvanishing gradients when training deep neural network models. For more on ReLU, see the tutorial:
The model predicts the probability of class 1 and uses the sigmoid activation function. The model is optimized using stochastic gradient descent and seeks to minimize thebinary cross-entropy loss.
The complete example is listed below.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
# pytorch mlp for binary classification
fromnumpyimportvstack
frompandasimportread_csv
fromsklearn.preprocessingimportLabelEncoder
fromsklearn.metricsimportaccuracy_score
fromtorch.utils.dataimportDataset
fromtorch.utils.dataimportDataLoader
fromtorch.utils.dataimportrandom_split
fromtorchimportTensor
fromtorch.nnimportLinear
fromtorch.nnimportReLU
fromtorch.nnimportSigmoid
fromtorch.nnimportModule
fromtorch.optimimportSGD
fromtorch.nnimportBCELoss
fromtorch.nn.initimportkaiming_uniform_
fromtorch.nn.initimportxavier_uniform_
# dataset definition
classCSVDataset(Dataset):
# load the dataset
def__init__(self,path):
# load the csv file as a dataframe
df=read_csv(path,header=None)
# store the inputs and outputs
self.X=df.values[:,:-1]
self.y=df.values[:,-1]
# ensure input data is floats
self.X=self.X.astype('float32')
# label encode target and ensure the values are floats
Running the example first reports the shape of the train and test datasets, then fits the model and evaluates it on the test dataset. Finally, a prediction is made for a single row of data.
Note: Yourresults may vary given the stochastic nature of the algorithm or evaluation procedure, or differences in numerical precision. Consider running the example a few times and compare the average outcome.
Further Reading
This section provides more resources on the topic if you are looking to go deeper.
It providesself-study tutorials withhundreds of working code to turn you from a novice to expert. It equips you with tensor operation,training,evaluation,hyperparameter optimization, and much more...
Kick-start your deep learning journey with hands-on exercises
Keras API is still looking simpler than Pytorch API, but Tensorflow’s version 2 is relatively more powerful and production-ready than Pytorch. Any reason behind in creating this tutorial or any plan/reason in switching to Pytorch?
I did evaluate Pytorch more than a year ago but did not switch finally. Another one was Julia but also sticking with TF and Keras until now, especially after having version 2 eager mode.
Sir a simple program I am running binary classification using DNN with pytorch. A discussion also going on in pytorch forum which is not yet solved. Can you help out I will post the link
Thank you for your tutorial. It is very nice. However, did you have any tutorial like this for tensorflow? Or from other source (book, web…)? Thank you!
My meant is: did you have any tutorial like this for only tensorflow without keras? Furthermore, I really like “prepare dataset” in your pytorch tutorial because I can be easy to customize my data set. However, this tutorial is not. It is instead of function in library that is difficult to customize. Can you explain why you do that? Thank you!
This is an amazing tutorial. Your end-to-end examples with very clear explanations take the prize! Thank you very much!
My favorite tutorial on your site is “How to Use the Keras Functional API for Deep Learning”, especially its part 5 “Multiple Input and Output Models”. That Keras tutorial helped me to develop far more sophisticated models.
While you’ve mentioned that you have no current plans to expand on PyTorch tutorials, it would be fantastic to see a similar PyTorch tutorial covering different multiple input/multiple output cases sometime in the future.
Hi Jason, Pytorch may provide some flexibility wrt Keras in deep learning issues according to my recent experiences i.e You can have deep insight what the model does and enhance somehow your coding skills. We look forward to new pytorch tutorials.
Dear Dr Jason, I thought I’d share this with you. For those with difficulties with the installation of Pytorch, I present another way of installing
When you thought that by pipping pytorch and its cuda and non-cuda variations seemed to be ok, you may get errors on either installation or DLL errors when trying to import from within python.
The following installation was straightforward and produced no errors.
If you one is having trouble with the PyTorch Installation particularly when you get errors during the installation, try the following:
I still do not understand why we extend Dataset when we create a custom dataset class, I know Dataset is an abstract class but what will happen if we will not extend it ??
Thank you so much for your great post. Recently, more and more new models are written in Pytorch. However, I am familiar with Keras. Do you have any ideas to transfer pytorch models to keras models? Many thanks
As always very useful tutorial Jason.Thanks. Am I not wrong in assuming that pytorch is more useful for people who are looking for complete control over their model i.e researchers.Having said that, keras does almost all my stuff.
Perhaps. Both tensorflow and pytorch give complete control, but for those interested in control, pytorch appears more popular – e.g. academics/researchers developing new methods rather than engineers solving problems.
Thank you for a great post. I am just curious: what do the numbers that go with MLP mean? For example MLP(13) and MLP(1). I see that you use different numbers in different examples.
Actually, I took the as is from your site, and just ran it. If I expand the predictions to: row = [0.00632,18.00,2.310,0,0.5380,6.5750,65.20,4.0900,1,296.0,15.30,396.90,4.98] yhat = predict(row, model) print(‘Predicted: %.3f’ % yhat) # should be near 24.00
row2 = [0.17505,0.00,5.960,0,0.4990,5.9660,30.20,3.8473,5,279.0,19.20,393.43,10.13] yhat2 = predict(row2, model) print(‘Predicted: %.3f’ % yhat2) #should be near 24.70
row3 = [2.77974,0.00,19.580,0,0.8710,4.9030,97.80,1.3459,5,403.0,14.70,396.90,29.29] yhat3 = predict(row3, model) print(‘Predicted: %.3f’ % yhat3) #should be near 11.80
row4 = [0.07503,33.00,2.180,0,0.4720,7.4200,71.90,3.0992,7,222.0,18.40,396.90,6.47] yhat4 = predict(row4, model) print(‘Predicted: %.3f’ % yhat4) #should be near 33.40
The SGD optimizer was getting stuck at a local minimum, changing it for the Adam optimizer works a lot better and you’ll see a noticeable response to different inputs.
I just tried this using the provided MLP regression code/data and got the same result as Hoang: All 4 test cases outputting the same value. If I plot all the predicted vs. actual in the test set, every output for the prediction is the same value.
I did the following modification to sanity check: I removed layers 2 and 3, and I just kept a single linear layer with the xavier uniform weight initialization, and it seems like I’m able to get an RMSE of 6.5 and with predictions on par with the actual for Hoang’s 4 cases. Plotting the predicted and actual in the test set also gave similar curves.
My hypothesis is: This just means that the Boston housing data is pretty much linear and adding multiple layers overfit the data? Could someone else corroborate this test and hypothesis?
BTW Jason, great site you got going here, I’m a software engineer too and starting to learn ML and I love your content. Keep up the great work, you’re doing a blessing to all devs wanting to get in the ML train ride!
I plotted actuals Vs predictions for the Regression example and see a constant line. Tried changing learning rate, epochs, Relu, number of hidden layers but did not help. This is true for other data sets aswell, not just Boston Housing dataset. Looks like something else needs to be changed in the program, any ideas.
The code is unmodified just added a plot to evaluate_model function, pl see below. My torch version is 1.5.1 running on anaconda windows 10. All code is run from command line. Could you check once on your system, if you are seeing a constant line for predictions.
Thanks Hemanth
def evaluate_model(test_dl, model): predictions, actuals = list(), list() for i, (inputs, targets) in enumerate(test_dl): # evaluate the model on the test set yhat = model(inputs) # retrieve numpy array yhat = yhat.detach().numpy() actual = targets.numpy() actual = actual.reshape((len(actual), 1)) # store predictions.append(yhat) actuals.append(actual) predictions, actuals = vstack(predictions), vstack(actuals) # calculate mse mse = mean_squared_error(actuals, predictions) pyplot.plot(actuals) pyplot.plot(predictions, color=’red’) pyplot.show() return mse
I was able to crack it even though it had Categorical Variables, however i couldn’t create a “prepare data” function as the Model requires many inputs:
class TabularModel(nn.Module): def __init__(self, embedding_size, num_numerical_cols, output_size, layers, p=0.4):
Can you please suggest how to write the “Preapre data” when there are categorical and numerical values ?
I am truly sorry for instantly jumping into questioning. The previous problem has disappeared now and dealing with another problem in the same program.
I got the same error as Bunga, only for Multi-class code. Can you help?
RuntimeError Traceback (most recent call last) in 149 model = MLP(4) 150 # train the model –> 151 train_model(train_dl, model) 152 # evaluate the model 153 acc = evaluate_model(test_dl, model)
in train_model(train_dl, model) 104 yhat = model(inputs) 105 # calculate loss –> 106 loss = criterion(yhat, targets) 107 # credit assignment 108 loss.backward()
the problem is targets var must be long() so change the targets variable at loss = criterion(yhat, targets) to loss = criterion(yhat, targets.long())
# train the model def train_model(train_dl, model): # define the optimization criterion = CrossEntropyLoss() optimizer = SGD(model.parameters(), lr=0.01, momentum=0.9) # enumerate epochs for epoch in range(500): # enumerate mini batches for i, (inputs, targets) in enumerate(train_dl): # clear the gradients optimizer.zero_grad() # compute the model output yhat = model(inputs) # calculate loss and change targets to long() loss = criterion(yhat, targets.long()) # credit assignment loss.backward() # update model weights optimizer.step()
File “C:\Users\haide\OneDrive\바탕 화면\temp.py”, line 150, in train_model(train_dl, model)
File “C:\Users\haide\OneDrive\바탕 화면\temp.py”, line 105, in train_model loss = criterion(yhat, targets)
File “C:\Users\haide\anaconda3\lib\site-packages\torch\nn\modules\module.py”, line 532, in __call__ result = self.forward(*input, **kwargs)
File “C:\Users\haide\anaconda3\lib\site-packages\torch\nn\modules\loss.py”, line 915, in forward return F.cross_entropy(input, target, weight=self.weight,
File “C:\Users\haide\anaconda3\lib\site-packages\torch\nn\functional.py”, line 2021, in cross_entropy return nll_loss(log_softmax(input, 1), target, weight, None, ignore_index, None, reduction)
File “C:\Users\haide\anaconda3\lib\site-packages\torch\nn\functional.py”, line 1838, in nll_loss ret = torch._C._nn.nll_loss(input, target, weight, _Reduction.get_enum(reduction), ignore_index)
RuntimeError: expected scalar type Long but found Int
Great tutorial, very clear. I run the MLP for Regression and when I change the predictor values of the “row” I get always the same that. Is it possible that the model is predicting only one value? Thanks
I am soon enrolling a course on DL at my work, now doing some exercises in advance to better understand the logic of e.g. perceptrons. Running your MLP for MC classification, I get the error shown below (just from copy’n’paste into my IDE). What am I doing wrong? I get the 100, 50, but accuracy info is blocked by the error happening.
Traceback (most recent call last): File “C:/Users/jcst/PycharmProjects/Deep_Learning_Projects/MLP_for_Multiclass_Classification.py”, line 166, in train_model(train_dl, model) File “C:/Users/jcst/PycharmProjects/Deep_Learning_Projects/MLP_for_Multiclass_Classification.py”, line 121, in train_model loss = criterion(yhat, targets) File “C:\Users\jcst\PycharmProjects\Deep_Learning_Projects\venv\lib\site-packages\torch\nn\modules\module.py”, line 722, in _call_impl result = self.forward(*input, **kwargs) File “C:\Users\jcst\PycharmProjects\Deep_Learning_Projects\venv\lib\site-packages\torch\nn\modules\loss.py”, line 948, in forward ignore_index=self.ignore_index, reduction=self.reduction) File “C:\Users\jcst\PycharmProjects\Deep_Learning_Projects\venv\lib\site-packages\torch\nn\functional.py”, line 2422, in cross_entropy return nll_loss(log_softmax(input, 1), target, weight, None, ignore_index, None, reduction) File “C:\Users\jcst\PycharmProjects\Deep_Learning_Projects\venv\lib\site-packages\torch\nn\functional.py”, line 2218, in nll_loss ret = torch._C._nn.nll_loss(input, target, weight, _Reduction.get_enum(reduction), ignore_index) RuntimeError: expected scalar type Long but found Int
Hi JC, You might want to try this on Google Colab. I got the same message (“RuntimeError: expected scalar type Long but found Int”) when I tried on my Jupiter notebook, but it worked on Google Colab.
I am always thankful for your valuable articles like this!
Just one minor comment for the loss function: The cross entropy loss (i.e., torch.nn.CrossEntropyLoss) in PyTorch combines nn.LogSoftmax() and nn.NLLLoss() in one single class, which means that you don’t need the softmax function in the case of multiclass classification problem (e.g., see the comments from ‘Oli’, ‘god_sp33d’, and so on). In fact, many PyTorch tutorials on multiclass classification do not use the softmax function for the said reason.
Jason: I am learning PyTorch through an on-line academic course and found your tutorial immensely helpful (as are all of your books). Please consider expanding your library with one or more Deep Learning books focused on PyTorch. You will find a very receptive audience.
Sure, it is a for loop, we are enumerating “train_dl” and each iteration we get a temp variable i for the iteration number and a tuple with the inputs and targets retrieved as the i’th item from “train_dl”.
Because the model is different. The LabelEncoder always outputs a vector of values, but if the model requires a matrix, you need to reshape a vector into a Nx1 matrix.
That depends on how the model designed, and the library used. It is better to consult the documentation of the functions you used. But I believe more likely, a matrix.
What is the “path” argument in __init__(self, path)?
I’m lost on where you’re actually importing the data. I see the read_csv line, but how does python know what data to import here when all you have written is “path”?
I’m wondering because I’m applying this tutorial to my own data, and when I import the data with read_csv using the actual file location (C:/Users/…), the argument “path” in __init__(self, path) is unused and I’m confused as to why it’s even there.
I also tried putting the actual file location in place of the argument “path” in __init__(self, path), but it underlines it and says “formal argument expected.”
Oh wow — nevermind, I did not see the path definition at the bottom of the script. I’m used to MATLAB where everything needs to be defined in a certain order.
You may be working on a regression problem and achieve zero prediction errors.
Alternately, you may be working on a classification problem and achieve 100% accuracy.
This is unusual and there are many possible reasons for this, including:
You are evaluating model performance on the training set by accident. Your hold out dataset (train or validation) is too small or unrepresentative. You have introduced a bug into your code and it is doing something different from what you expect. Your prediction problem is easy or trivial and may not require machine learning. The most common reason is that your hold out dataset is too small or not representative of the broader problem.
This can be addressed by:
Using k-fold cross-validation to estimate model performance instead of a train/test split. Gather more data. Use a different split of data for train and test, such as 50/50.
Hi Marilyn…We do not promote any particular OS or Python environment. Are you having issues with the code running such that we may help provide suggestions?
Hello Jason, First, thank you so much for this great post! I love your posts, it is a great help for me. This is however my first writing to you.
I have a question w.r.t. this line: model = MLP(34). I understand that 34 is the input for the first layer, right? But how did you determine that it should be 34? I did not understand that.
Furthermore, when I tried using other numbers, I got an error. For example, doing MLP(33), I get the following error: “RuntimeError: mat1 and mat2 shapes cannot be multiplied (32×34 and 33×10)”
Hi Nik…You are very welcome! You’re correct that the34 inmodel = MLP(34) represents the number of input features for the first layer of the Multi-Layer Perceptron (MLP). This value is derived directly from the number of features (columns) in your dataset, excluding the target (label) column.
### Why 34? In your code, the dataset is loaded from the CSV file using this line:
python df = read_csv(path, header=None)
This loads all the columns in the CSV file, including the target column. The code then separates the features and the target like this:
python self.X = df.values[:, :-1] # Features (all columns except the last one) self.y = df.values[:, -1] # Target (the last column)
–self.X contains all the columns except the last one, meaning these are the features used for the model input. – In theionosphere.csv dataset you’re using, there are 35 columns in total (based on how the data is structured), where 34 are the input features (columns) and the 35th is the label (target). Hence, the input layer of the model expects 34 features, so you need to pass34 as the argument to the MLP when initializing it:MLP(34).
### Error with Other Input Sizes When you use a different number, likeMLP(33), you get the following error:
RuntimeError: mat1 and mat2 shapes cannot be multiplied (32×34 and 33×10)
This error occurs because the input shape to the network does not match the number of features in your dataset. Specifically, the first dimension of the input tensor (which should match the number of features) is expected to be 34, but you’re passing a number that doesn’t match the actual feature count.
In PyTorch, matrix multiplication (which happens during forward propagation) requires that the number of columns in the input tensor matches the number of rows in the weight matrix. If there’s a mismatch, as in your case when you pass 33 instead of 34, the operation fails because the dimensions don’t align.
### Solution: You need to ensure that the input to the MLP matches the number of features in your dataset. Since the dataset you’re using has 34 input features, the input to the first layer should be34 in theMLP class:
python model = MLP(34)
If you change the input feature count, you also need to adjust the dataset to match the new number of features accordingly.
Let me know if you’d like further clarification or have additional questions!
