File size: 4,993 Bytes
33b03a3 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 |
import matplotlib.pyplot as plt
import numpy as np
font_size = 14
def fid_scaling_law_no_cfg():
# data
steps = np.array([50, 100, 200, 300,])
loss_b = np.array([41.025, 33.442, 32.105, 32.196])
loss_l = np.array([25.889, 24.654, 19.742, 19.070])
loss_xl = np.array([19.820, 18.037, 14.772, 15.549])
steps_ = np.array([50, 200, 300,])
loss_xxl = np.array([17.195, 13.997, 14.648])
loss_3b = np.array([16.431, 9.949, 9.380])
# Plot
plt.figure(figsize=(6, 4))
plt.plot(steps, loss_b, 'o-', label='B', color='red')
plt.plot(steps, loss_l, 'o-', label='L', color='orange')
plt.plot(steps, loss_xl, 'o-', label='XL', color='green')
plt.plot(steps_, loss_xxl, 'o-', label='XXL', color='blue')
plt.plot(steps_, loss_3b, 'o-', label='3B', color='purple')
plt.xlabel('Training Epochs', fontsize=font_size)
plt.ylabel('FID', fontsize=font_size)
# plt.grid(True)
# plt.yscale('log')
# Customize the plot to match the appearance of the provided figure
plt.legend(loc='upper right', framealpha=0.5, fontsize=font_size, facecolor='white')
# Customizing the x and y axis ticks (to match the example's steps)
# plt.xticks(np.linspace(0, 800000, 5), ['0', '200K', '400K', '600K', '800K'])
plt.yticks(np.arange(5, 50, step=5))
# Show plot
plt.tight_layout()
plt.savefig('fid_scaling_law_no_cfg.png', dpi=600)
def fid_scaling_law_cfg():
# data
steps = np.array([50, 100, 200, 300,])
loss_b_cfg = np.array([8.309, 7.256, 6.542, 6.249])
loss_l_cfg = np.array([4.240, 3.705, 3.220, 3.075])
loss_xl_cfg = np.array([3.420, 3.089, 2.617, 2.629])
steps_ = np.array([50, 200, 300,])
loss_xxl_cfg = np.array([2.893, 2.331, 2.340])
loss_3b_cfg = np.array([2.611, 2.381, 2.329])
# Plot
plt.figure(figsize=(6, 4))
plt.plot(steps, loss_b_cfg, 'o-', label='B', color='red')
plt.plot(steps, loss_l_cfg, 'o-', label='L', color='orange')
plt.plot(steps, loss_xl_cfg, 'o-', label='XL', color='green')
plt.plot(steps_, loss_xxl_cfg, 'o-', label='XXL', color='blue')
plt.plot(steps_, loss_3b_cfg, 'o-', label='3B', color='purple')
plt.xlabel('Training Epochs', fontsize=font_size)
plt.ylabel('FID', fontsize=font_size)
# plt.grid(True)
# plt.yscale('log')
# Customize the plot to match the appearance of the provided figure
plt.legend(loc='upper right', framealpha=0.5, fontsize=font_size, facecolor='white')
# Customizing the x and y axis ticks (to match the example's steps)
# plt.xticks(np.linspace(0, 800000, 5), ['0', '200K', '400K', '600K', '800K'])
plt.yticks(np.arange(2, 9, step=1))
# Show plot
plt.tight_layout()
plt.savefig('fid_scaling_law_cfg.png', dpi=600)
def sample_topk():
# Data
top_k = np.array([16384, 10000, 8000, 6000, 4000, 2000, 1000])
fid_values = np.array([3.075, 3.369, 3.643, 3.969, 4.635, 5.998, 7.428])
inception_scores = np.array([256.067, 265.222, 268.237, 270.159, 271.455, 267.278, 251.268])
fig, ax1 = plt.subplots()
# Create first y-axis
ax1.set_xlabel('top-k', fontsize=font_size)
ax1.set_ylabel('FID', color='teal', fontsize=font_size)
ax1.plot(top_k, fid_values, 'o-', color='teal', label="FID")
ax1.tick_params(axis='y', labelcolor='teal')
ax1.tick_params(axis='x')
# Create second y-axis
ax2 = ax1.twinx()
ax2.set_ylabel('Inception Score', color='brown', fontsize=font_size)
ax2.plot(top_k, inception_scores, 'o-', color='brown', label="Inception Score")
ax2.tick_params(axis='y', labelcolor='brown')
# Adding a legend
fig.legend(loc='upper right', bbox_to_anchor=(1.0, 1.0), bbox_transform=ax1.transAxes, fontsize=font_size)
fig.tight_layout() # Adjust layout to prevent overlap
plt.savefig('effect_topk.png', dpi=600)
def sample_cfg():
# Data
cfg = np.array([1.5, 1.75, 2.00, 2.25])
fid_values = np.array([4.743, 3.151, 3.075, 3.620])
inception_scores = np.array([165.381, 214.152, 256.067, 291.695])
plt.figure(figsize=(10, 4))
fig, ax1 = plt.subplots()
# Create first y-axis
ax1.set_xlabel('cfg', fontsize=font_size)
ax1.set_ylabel('FID', color='teal', fontsize=font_size)
ax1.plot(cfg, fid_values, 'o-', color='teal', label="FID")
ax1.tick_params(axis='y', labelcolor='teal')
ax1.tick_params(axis='x')
# Create second y-axis
ax2 = ax1.twinx()
ax2.set_ylabel('Inception Score', color='brown', fontsize=font_size)
ax2.plot(cfg, inception_scores, 'o-', color='brown', label="Inception Score")
ax2.tick_params(axis='y', labelcolor='brown')
# Adding a legend
fig.legend(loc='upper right', bbox_to_anchor=(1.0, 1.0), bbox_transform=ax1.transAxes, fontsize=font_size)
fig.tight_layout() # Adjust layout to prevent overlap
plt.savefig('effect_cfg.png', dpi=600)
if __name__ == "__main__":
fid_scaling_law_no_cfg()
fid_scaling_law_cfg()
sample_cfg()
sample_topk()
|