CLIP：用对比学习对齐图像和文本

logits = I @ T.T / temperature

import math


def normalize(vector):
    length = math.sqrt(sum(value * value for value in vector))
    if length == 0:
        raise ValueError("零向量不能归一化")
    return [value / length for value in vector]


def cosine(left, right):
    return sum(a * b for a, b in zip(normalize(left), normalize(right)))


image = [0.90, 0.10, 0.00]  # 示例视觉编码，不是手工像素
texts = {
    "设备指示灯为绿色": [0.85, 0.15, 0.00],
    "设备指示灯为红色": [0.05, 0.90, 0.05],
    "一只动物": [0.00, 0.10, 0.90],
}
scores = {label: round(cosine(image, vector), 4) for label, vector in texts.items()}
print(scores)
print(max(scores, key=scores.get))