Tensor creation (torch.tensor, zeros, ones, rand) in PyTorch - Model Pipeline Trace

Model Pipeline - Tensor creation (torch.tensor, zeros, ones, rand)

This pipeline shows how tensors are created using different PyTorch functions. Tensors are the basic building blocks for data in PyTorch, similar to arrays or tables in everyday life.

Data Flow - 4 Stages

1Create tensor from list

N/A→torch.tensor creates a tensor from a Python list→3 rows x 3 columns

[[1, 2, 3], [4, 5, 6], [7, 8, 9]] -> torch.tensor -> tensor([[1, 2, 3], [4, 5, 6], [7, 8, 9]])

↓

2Create tensor of zeros

N/A→torch.zeros creates a tensor filled with zeros of given shape→2 rows x 4 columns

torch.zeros(2, 4) -> tensor([[0., 0., 0., 0.], [0., 0., 0., 0.]])

↓

3Create tensor of ones

N/A→torch.ones creates a tensor filled with ones of given shape→3 rows x 2 columns

torch.ones(3, 2) -> tensor([[1., 1.], [1., 1.], [1., 1.]])

↓

4Create tensor with random values

N/A→torch.rand creates a tensor with random values between 0 and 1→2 rows x 3 columns

torch.rand(2, 3) -> tensor([[0.23, 0.67, 0.12], [0.89, 0.45, 0.33]])

Training Trace - Epoch by Epoch

No training loss to show because this pipeline only creates tensors.

Epoch	Loss ↓	Accuracy ↑	Observation
1	N/A	N/A	No training occurs because this is tensor creation only.

Prediction Trace - 4 Layers

Layer 1: torch.tensor

Layer 2: torch.zeros

Layer 3: torch.ones

Layer 4: torch.rand

Model Quiz - 3 Questions

What does torch.zeros(3, 2) produce?

A A tensor with 2 rows and 3 columns filled with zeros

B A tensor with 3 rows and 2 columns filled with ones

C A tensor with 3 rows and 2 columns filled with zeros

D A tensor with random values between 0 and 1

Key Insight

Tensor creation functions in PyTorch provide flexible ways to start working with data. Knowing how to create tensors filled with zeros, ones, or random values is essential for initializing models and data structures before training.