ML Model Inference Time KDE

import numpy as np
import matplotlib.pyplot as plt
from scipy.stats import gaussian_kde

np.random.seed(333)
BG_COLOR = '#0a0a0f'
TEXT_COLOR = 'white'

# Inference time in ms
cloud_gpu = np.random.lognormal(2.5, 0.4, 500)
edge_device = np.random.lognormal(3.5, 0.5, 500)
mobile = np.random.lognormal(4.2, 0.6, 500)

fig, ax = plt.subplots(figsize=(10, 6), facecolor=BG_COLOR)
ax.set_facecolor(BG_COLOR)

x_range = np.linspace(0, 300, 500)

for data, color, label in [(cloud_gpu, '#4927F5', 'Cloud GPU'), 
                            (edge_device, '#27F5B0', 'Edge Device'),
                            (mobile, '#F5B027', 'Mobile')]:
    data_clipped = data[data < 300]
    kde = gaussian_kde(data_clipped)
    density = kde(x_range)
    ax.plot(x_range, density, color=color, linewidth=2.5, label=label)
    ax.fill_between(x_range, density, alpha=0.3, color=color)

ax.axvline(100, color='#F5276C', linestyle='--', alpha=0.8, linewidth=2, label='Real-time: 100ms')

ax.set_xlabel('Inference Time (ms)', fontsize=12, color=TEXT_COLOR, fontweight='500')
ax.set_ylabel('Density', fontsize=12, color=TEXT_COLOR, fontweight='500')
ax.set_title('ML Model Inference Latency by Platform', fontsize=14, color=TEXT_COLOR, fontweight='bold', pad=15)
ax.set_xlim(0, 300)

ax.tick_params(colors='#888888', labelsize=10)
for spine in ax.spines.values():
    spine.set_color('#333333')
ax.legend(facecolor=BG_COLOR, edgecolor='#333333', labelcolor=TEXT_COLOR, fontsize=10)

plt.tight_layout()
plt.show()