Common Python Mistakes & Fixes

A quick reference set covering common mistakes in Python coding practice, including tensor definitions, bias usage, and handling multiple conditions.

What is the correct way to define a tensor with torch.tensor?

Use a tuple with a single argument, enclosed in brackets ( ).

What common mistake happens when defining a tensor?

Forgetting that torch.tensor expects a tuple with just one argument.

How should you check for bias in a class method?

Use if self.bias: — always include self for class functions.

What is the role of self in Python class methods?

It ensures the method has access to the instance’s attributes and functions.

What should you be careful of when using multiple and and or statements?

Verify all possible cases and ensure logic is grouped correctly.

What happens if one of the conditions in a complex and/or chain isn’t satisfied?

The result may not behave as expected — test each case explicitly.

What is the shape of the weight matrix

(out_features, in_features)

How do you define self.weights

self.weights = nn.Parameters(torch.empty(out_featuree, in_features))

Bias parameter

self.register_parameter(“bias”, None)

Initializjing weights

nn.init.xavier_uniform(self.weight)

Matrix Multiplication

torch.nn.functional.linear(x, self.weight, self.bias)

What are Modules and Parameters

Modules is a nueral network

Parameter is weight for one layer.

Relu

torch.clamp(x, min=0)

Tanx

torch.ones_like(x)/ (torch.oneslike(x) + torch.exp(-x))

Leaky Relu

torch.where(x>=0, x, x*negative_slope)

Gradient is negative slope

What is channels in computer vision

self.channels = channels in the dataset that represent different features or color depths of the images, typically corresponding to RGB in color images.

How do you increase number of channels

Run multiple number of kernels in parallel.

Paramerters for kernel

stride, padding, kernel_size

What does x.shape for for compute rvision

batch_size, in_channels, input_h, input_w

Padding

x = F.pad(x,(padding[1],padding[1], padding[0]. padding[0]])

Symbol for integer multiplication

//

What does a*b do for tensors

It deoes elementwise multiplicaiton and not mtrix multiplixation.

How do you get sum of all entires in a matrix

torch.sum(a)

Max Pooling`

Runs a max kernel over the input. The max value is taken from each region covered by the kernel, reducing the dimensions of the input.