📘

Temperature Sampler

Given a list of logits (raw scores) for vocabulary words and a temperature value, apply temperature scaling and compute the resulting probability distribution using softmax.

Temperature scaling divides each logit by the temperature before applying softmax: $p_i = \frac{e^{z_i / T}}{\sum_j e^{z_j / T}}$

Input format:

Line 1: Comma-separated word:logit pairs (e.g., "cat:2.0,dog:1.0,fish:0.5")
Line 2: Temperature T (float)

Output: Each word and its probability (rounded to 4 decimal places), sorted by probability descending, one per line as "word probability".

Example:

Input:

cat:2.0,dog:1.0,fish:0.5
1.0

Output:

cat 0.6285
dog 0.2312
fish 0.1402

Reasoning:

Step 1: Divide logits by temperature (T=1.0) cat: 2.0/1.0 = 2.0, dog: 1.0/1.0 = 1.0, fish: 0.5/1.0 = 0.5

Step 2: Apply softmax e^2.0 ≈ 7.389, e^1.0 ≈ 2.718, e^0.5 ≈ 1.649 Sum ≈ 11.756

Step 3: Normalize cat: 7.389/11.756 ≈ 0.6285 dog: 2.718/11.756 ≈ 0.2312 fish: 1.649/11.756 ≈ 0.1402

Constraints:

Temperature T > 0
Apply softmax after dividing logits by T
Round probabilities to 4 decimal places
Output sorted by probability descending (ties: alphabetical)

Editor

Python 3.13.1

Test Results

0/0

Run code to see test results.

📘