Indexing and slicing in PyTorch - Model Pipeline Trace

Model Pipeline - Indexing and slicing

This pipeline shows how data tensors are accessed and manipulated using indexing and slicing in PyTorch. It helps select parts of data for further processing or model input.

Data Flow - 4 Stages

1Input tensor

1 tensor of shape (4, 5)→Create a 2D tensor with 4 rows and 5 columns→1 tensor of shape (4, 5)

[[10, 20, 30, 40, 50], [60, 70, 80, 90, 100], [110, 120, 130, 140, 150], [160, 170, 180, 190, 200]]

↓

2Row indexing

1 tensor of shape (4, 5)→Select the 2nd row using indexing→1 tensor of shape (5,)

[60, 70, 80, 90, 100]

↓

3Column slicing

1 tensor of shape (4, 5)→Select columns 1 to 3 (exclusive) for all rows using slicing→1 tensor of shape (4, 2)

[[20, 30], [70, 80], [120, 130], [170, 180]]

↓

4Combined indexing and slicing

1 tensor of shape (4, 5)→Select the last two columns of the 3rd row→1 tensor of shape (2,)

[140, 150]

Training Trace - Epoch by Epoch

Loss
1.0 | *       
0.8 |  *      
0.6 |   *     
0.4 |    *    
0.2 |     *   
0.0 +---------
      1 2 3 4 5
      Epochs

Epoch	Loss ↓	Accuracy ↑	Observation
1	0.85	0.40	Initial loss is high; accuracy is low as model starts learning.
2	0.65	0.55	Loss decreases and accuracy improves after first update.
3	0.50	0.70	Model learns important features; metrics improve steadily.
4	0.38	0.80	Loss continues to drop; accuracy reaches a good level.
5	0.30	0.85	Training converges with low loss and high accuracy.

Prediction Trace - 4 Layers

Layer 1: Input tensor

Layer 2: Row indexing

Layer 3: Column slicing

Layer 4: Combined indexing and slicing

Model Quiz - 3 Questions

What is the shape of the tensor after selecting the 2nd row from a (4, 5) tensor?

A (1, 5)

B (4,)

C (5,)

D (4, 5)

Key Insight

Indexing and slicing let us pick specific parts of data tensors easily. This is useful to focus on relevant features or samples before feeding data into a model. Understanding shapes after these operations helps avoid errors and ensures correct data flow.