Edge models — measure sparsity

Background

After pruning you want to know how sparse the model became: the fraction of weights that are exactly zero. Higher sparsity means a smaller, faster model (with the right kernels).

$$\text{sparsity} = \frac{\#\{w : w = 0\}}{\#\text{weights}}$$

Problem statement

Implement sparsity(W) returning the fraction of entries that are exactly 0.

Input

• W — array-like weight tensor.

Output

Returns a Python float in $[0, 1]$.

Examples

Example 1

Input:  W = [[0, 1], [2, 0]]
Output: 0.5

Explanation: 2 of the 4 entries are zero.

Example 2 — fully pruned

Input:  W = [0, 0, 0]
Output: 1.0

Constraints

• Count exact zeros divided by the total number of entries.
• Return a plain float.

Notes

• Reported alongside accuracy, sparsity tells you the size/speed payoff of pruning — e.g. "90% sparse with <1% accuracy drop".