Dataset from tensors in TensorFlow - Model Pipeline Trace

Model Pipeline - Dataset from tensors

This pipeline shows how to create a dataset directly from tensors, then use it to train a simple model. It starts with raw data in tensors, builds a dataset, trains a model, and makes predictions.

Data Flow - 3 Stages

1Raw tensors

None→Create tensors for features and labels→5 rows x 3 columns (features), 5 rows x 1 column (labels)

[[1.0, 2.0, 3.0], [4.0, 5.0, 6.0], [7.0, 8.0, 9.0], [10.0, 11.0, 12.0], [13.0, 14.0, 15.0]] (features), [0, 1, 0, 1, 0] (labels)

↓

2Dataset creation

5 rows x 3 columns (features), 5 rows x 1 column (labels)→Create tf.data.Dataset from tensors→5 elements of (3 features, 1 label)

Dataset element example: (features=[1.0, 2.0, 3.0], label=0)

↓

3Batching

5 elements of (3 features, 1 label)→Batch dataset into groups of 2→3 batches: 2 elements, 2 elements, 1 element

Batch 1: features=[[1.0,2.0,3.0],[4.0,5.0,6.0]], labels=[0,1]

Training Trace - Epoch by Epoch

Loss
0.7 |****
0.6 |*** 
0.5 |**  
0.4 |*   
0.3 |    
     1 2 3 4 5 Epochs

Epoch	Loss ↓	Accuracy ↑	Observation
1	0.693	0.40	Initial loss is high, accuracy low as model starts learning
2	0.580	0.60	Loss decreases, accuracy improves as model learns patterns
3	0.480	0.80	Model continues to improve, loss goes down, accuracy up
4	0.400	0.80	Loss decreases further, accuracy stabilizes
5	0.350	1.00	Model fits data well, perfect accuracy on training set

Prediction Trace - 3 Layers

Layer 1: Input layer

Layer 2: Dense layer with ReLU

Layer 3: Output layer with sigmoid

Model Quiz - 3 Questions

What is the shape of the dataset after batching?

A 3 batches: 2 elements, 2 elements, 1 element

B 5 batches: 1 element each

C 1 batch: 5 elements

D 2 batches: 3 elements, 2 elements

Key Insight

Creating a dataset from tensors allows easy feeding of data into TensorFlow models. Batching helps training efficiency. Loss decreasing and accuracy increasing show the model is learning well.