How to Get Model Parameters in PyTorch: Syntax and Examples

In PyTorch, you can get model parameters using the model.parameters() method, which returns an iterator over all parameters (weights and biases). To get parameter names along with values, use model.named_parameters().

Syntax

The main ways to get model parameters in PyTorch are:

• model.parameters(): Returns an iterator over all parameter tensors.
• model.named_parameters(): Returns an iterator over (name, parameter) pairs.

These methods help you access weights and biases inside your model.

python

parameters = model.parameters()
named_parameters = model.named_parameters()

Example

This example shows how to define a simple neural network and print its parameters using both parameters() and named_parameters().

python

import torch
import torch.nn as nn

class SimpleNet(nn.Module):
    def __init__(self):
        super().__init__()
        self.fc1 = nn.Linear(4, 3)
        self.fc2 = nn.Linear(3, 2)

    def forward(self, x):
        x = self.fc1(x)
        x = self.fc2(x)
        return x

model = SimpleNet()

# Get parameters only
print('Parameters:')
for param in model.parameters():
    print(param.shape)

# Get named parameters
print('\nNamed Parameters:')
for name, param in model.named_parameters():
    print(f'{name}: {param.shape}')

Output

Parameters:
torch.Size([3, 4])
torch.Size([3])
torch.Size([2, 3])
torch.Size([2])

Named Parameters:
fc1.weight: torch.Size([3, 4])
fc1.bias: torch.Size([3])
fc2.weight: torch.Size([2, 3])
fc2.bias: torch.Size([2])

Common Pitfalls

Common mistakes when getting model parameters include:

• Trying to access parameters before model initialization.
• Confusing parameters() (values only) with named_parameters() (names and values).
• Modifying parameters without using with torch.no_grad(), which can break gradient tracking.

python

import torch
import torch.nn as nn

model = nn.Linear(2, 1)

# Wrong: Trying to print parameter names using parameters() only
for param in model.parameters():
    print(param)  # No names here

# Right: Use named_parameters() to get names
for name, param in model.named_parameters():
    print(name, param.shape)

Output

tensor([[ 0.1234, -0.5678]], requires_grad=True)
tensor([0.4321], requires_grad=True)
weight torch.Size([1, 2])
bias torch.Size([1])

Quick Reference

Method	Description
model.parameters()	Iterator over all parameter tensors (weights, biases)
model.named_parameters()	Iterator over (name, parameter) pairs
param.data	Access raw tensor data of a parameter
param.requires_grad	Check if parameter will be updated during training

Key Takeaways

Use model.parameters() to get all parameter tensors in PyTorch models.

Use model.named_parameters() to get parameter names along with tensors.

Always ensure the model is initialized before accessing parameters.

Avoid modifying parameters directly without disabling gradient tracking.

Parameter shapes help understand model layer sizes and connections.