Float types (float16, float32, float64) in NumPy - Step-by-Step Execution

Concept Flow - Float types (float16, float32, float64)

Start: Choose float type

↓

Create numpy array with chosen float type

↓

Store numbers in array

↓

Check memory size per element

↓

Use array for calculations

↓

Observe precision and memory trade-off

↓

End

This flow shows how to create numpy arrays with different float types, check their memory size, and understand precision differences.

Execution Sample

NumPy

import numpy as np
arr16 = np.array([1.5, 2.5], dtype=np.float16)
arr32 = np.array([1.5, 2.5], dtype=np.float32)
arr64 = np.array([1.5, 2.5], dtype=np.float64)
print(arr16.itemsize, arr32.itemsize, arr64.itemsize)

Creates three numpy arrays with float16, float32, and float64 types and prints their memory size per element.

Execution Table

Step	Action	Array Created	Data Type	Item Size (bytes)	Output
1	Create array with float16	[1.5, 2.5]	float16	2	None
2	Create array with float32	[1.5, 2.5]	float32	4	None
3	Create array with float64	[1.5, 2.5]	float64	8	None
4	Print item sizes	N/A	N/A	N/A	2 4 8
5	End	N/A	N/A	N/A	Execution complete

💡 All arrays created and item sizes printed, showing memory usage differences.

Variable Tracker

Variable	Start	After Step 1	After Step 2	After Step 3	Final
arr16	undefined	[1.5, 2.5] (float16)	[1.5, 2.5] (float16)	[1.5, 2.5] (float16)	[1.5, 2.5] (float16)
arr32	undefined	undefined	[1.5, 2.5] (float32)	[1.5, 2.5] (float32)	[1.5, 2.5] (float32)
arr64	undefined	undefined	undefined	[1.5, 2.5] (float64)	[1.5, 2.5] (float64)

Key Moments - 2 Insights

Why does float16 use less memory than float64?

Because float16 uses 2 bytes per number while float64 uses 8 bytes, as shown in execution_table step 4 where item sizes are printed.

Does using float16 affect the precision of numbers?

Yes, float16 has less precision than float64, meaning it can store fewer decimal places accurately. This is implied by the different data types in execution_table steps 1-3.

Visual Quiz - 3 Questions

Look at the execution table at step 4, what is the item size in bytes for float32?

A 2

B 4

C 8

D 16

💡 Hint

Check the 'Item Size (bytes)' column in execution_table row for step 4.

At which step is the float64 array created?

A Step 3

B Step 2

C Step 1

D Step 4

💡 Hint

Look at the 'Action' and 'Data Type' columns in execution_table.

If you want to save memory but keep some decimal precision, which float type would you choose based on the execution?

A float64

B float16

C float32

D int32

💡 Hint

Consider the item sizes and typical precision trade-offs shown in execution_table and key_moments.

Concept Snapshot

Float types in numpy:
- float16 uses 2 bytes, less precision
- float32 uses 4 bytes, balanced precision
- float64 uses 8 bytes, highest precision
Choose based on memory vs precision needs
Use dtype parameter to set type

Full Transcript

This lesson shows how to create numpy arrays with different float types: float16, float32, and float64. Each type uses different memory sizes per number: 2, 4, and 8 bytes respectively. We create arrays with each type and print their item sizes to see the difference. Float16 uses less memory but has less precision, while float64 uses more memory but stores numbers more precisely. This helps decide which float type to use depending on your needs for memory and accuracy.