"""
直接生成 Baseline vs +Tricks 训练对比图（基于典型实验曲线）
不需要跑训练，立即可看效果
"""
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
import numpy as np
import os

# 模拟典型实验曲线
epochs = np.arange(16)

# Baseline: 快速过拟合，train↑ val 停滞后下降
base_train_loss = np.array([1.50,1.10,0.91,0.77,0.66,0.57,0.48,0.40,0.33,0.27,0.22,0.19,0.16,0.14,0.12,0.11])
base_val_loss   = np.array([1.34,1.00,0.91,0.80,0.84,0.75,0.80,0.85,0.92,1.00,1.08,1.15,1.22,1.28,1.33,1.38])
base_train_acc  = np.array([0.45,0.61,0.68,0.73,0.77,0.80,0.83,0.86,0.88,0.90,0.92,0.93,0.94,0.95,0.96,0.96])
base_val_acc    = np.array([0.53,0.65,0.68,0.72,0.71,0.74,0.73,0.72,0.71,0.70,0.69,0.68,0.67,0.66,0.65,0.64])

# +Tricks: 平稳收敛，train 和 val 差距小
tri_train_loss = np.array([1.55,1.15,0.95,0.82,0.72,0.64,0.58,0.52,0.47,0.43,0.39,0.36,0.33,0.30,0.28,0.26])
tri_val_loss   = np.array([1.38,1.05,0.88,0.78,0.71,0.65,0.60,0.56,0.53,0.50,0.48,0.46,0.45,0.44,0.43,0.42])
tri_train_acc  = np.array([0.43,0.59,0.66,0.71,0.75,0.78,0.81,0.83,0.85,0.87,0.88,0.89,0.90,0.91,0.92,0.92])
tri_val_acc    = np.array([0.51,0.63,0.69,0.73,0.76,0.78,0.80,0.81,0.82,0.83,0.84,0.84,0.85,0.85,0.86,0.86])

fig, axes = plt.subplots(1, 2, figsize=(14, 5))

# Loss
ax = axes[0]
ax.plot(epochs, base_train_loss, 'b-',  alpha=0.7, linewidth=2, label='Baseline train')
ax.plot(epochs, base_val_loss,   'b--', alpha=0.9, linewidth=2, label='Baseline val')
ax.plot(epochs, tri_train_loss,  'r-',  alpha=0.7, linewidth=2, label='+Tricks train')
ax.plot(epochs, tri_val_loss,    'r--', alpha=0.9, linewidth=2, label='+Tricks val')
ax.axvline(x=5, color='gray', linestyle=':', alpha=0.5, label='Early stop (Baseline)')
ax.set_xlabel('Epoch', fontsize=12)
ax.set_ylabel('Loss', fontsize=12)
ax.set_title('Loss: Baseline vs +Tricks', fontsize=13, fontweight='bold')
ax.legend(loc='upper right')
ax.grid(True, alpha=0.3)
ax.set_ylim(0, 1.8)

# Accuracy
ax = axes[1]
ax.plot(epochs, base_train_acc, 'b-',  alpha=0.7, linewidth=2)
ax.plot(epochs, base_val_acc,   'b--', alpha=0.9, linewidth=2)
ax.plot(epochs, tri_train_acc,  'r-',  alpha=0.7, linewidth=2)
ax.plot(epochs, tri_val_acc,    'r--', alpha=0.9, linewidth=2)
ax.axvline(x=5, color='gray', linestyle=':', alpha=0.5)
ax.set_xlabel('Epoch', fontsize=12)
ax.set_ylabel('Accuracy', fontsize=12)
ax.set_title('Accuracy: Baseline vs +Tricks', fontsize=13, fontweight='bold')
ax.legend(['Baseline train','Baseline val','+Tricks train','+Tricks val','Early stop (Baseline)'], loc='lower right')
ax.grid(True, alpha=0.3)
ax.set_ylim(0.4, 1.0)

# 添加注释
axes[0].annotate('过拟合！\ntrain↓ val↑', xy=(10, 1.15), fontsize=11, color='blue',
                ha='center', bbox=dict(boxstyle='round', facecolor='lightblue', alpha=0.5))
axes[1].annotate('过拟合！\ntrain↑ val↓', xy=(10, 0.70), fontsize=11, color='blue',
                ha='center', bbox=dict(boxstyle='round', facecolor='lightblue', alpha=0.5))
axes[0].annotate('平稳收敛\ngap 小', xy=(12, 0.50), fontsize=11, color='red',
                ha='center', bbox=dict(boxstyle='round', facecolor='lightyellow', alpha=0.5))
axes[1].annotate('平稳提升\ngap 小', xy=(12, 0.82), fontsize=11, color='red',
                ha='center', bbox=dict(boxstyle='round', facecolor='lightyellow', alpha=0.5))

plt.tight_layout()
out_path = os.path.join(os.path.dirname(__file__), 'training_tricks_comparison.png')
plt.savefig(out_path, dpi=150, bbox_inches='tight')
print(f"✅ 对比图已保存: {out_path}")
plt.close()

# 同时保存到 copyparty 目录
copyparty_path = '/root/copyparty-files/deeplearning/code/training_tricks_comparison.png'
plt.figure(figsize=(14, 5))
fig2, axes2 = plt.subplots(1, 2, figsize=(14, 5))
ax = axes2[0]
ax.plot(epochs, base_train_loss, 'b-',  alpha=0.7, linewidth=2, label='Baseline train')
ax.plot(epochs, base_val_loss,   'b--', alpha=0.9, linewidth=2, label='Baseline val')
ax.plot(epochs, tri_train_loss,  'r-',  alpha=0.7, linewidth=2, label='+Tricks train')
ax.plot(epochs, tri_val_loss,    'r--', alpha=0.9, linewidth=2, label='+Tricks val')
ax.axvline(x=5, color='gray', linestyle=':', alpha=0.5)
ax.set_xlabel('Epoch'); ax.set_ylabel('Loss'); ax.set_title('Loss: Baseline vs +Tricks', fontweight='bold')
ax.legend(); ax.grid(True, alpha=0.3); ax.set_ylim(0, 1.8)
ax.annotate('过拟合！\ntrain↓ val↑', xy=(10, 1.15), fontsize=11, color='blue',
           ha='center', bbox=dict(boxstyle='round', facecolor='lightblue', alpha=0.5))
ax.annotate('平稳收敛\ngap 小', xy=(12, 0.50), fontsize=11, color='red',
           ha='center', bbox=dict(boxstyle='round', facecolor='lightyellow', alpha=0.5))

ax = axes2[1]
ax.plot(epochs, base_train_acc, 'b-',  alpha=0.7, linewidth=2)
ax.plot(epochs, base_val_acc,   'b--', alpha=0.9, linewidth=2)
ax.plot(epochs, tri_train_acc,  'r-',  alpha=0.7, linewidth=2)
ax.plot(epochs, tri_val_acc,    'r--', alpha=0.9, linewidth=2)
ax.axvline(x=5, color='gray', linestyle=':', alpha=0.5)
ax.set_xlabel('Epoch'); ax.set_ylabel('Accuracy'); ax.set_title('Accuracy: Baseline vs +Tricks', fontweight='bold')
ax.legend(['Baseline train','Baseline val','+Tricks train','+Tricks val','Early stop'], loc='lower right')
ax.grid(True, alpha=0.3); ax.set_ylim(0.4, 1.0)
ax.annotate('过拟合！\ntrain↑ val↓', xy=(10, 0.70), fontsize=11, color='blue',
           ha='center', bbox=dict(boxstyle='round', facecolor='lightblue', alpha=0.5))
ax.annotate('平稳提升\ngap 小', xy=(12, 0.82), fontsize=11, color='red',
           ha='center', bbox=dict(boxstyle='round', facecolor='lightyellow', alpha=0.5))
plt.tight_layout()
plt.savefig(copyparty_path, dpi=150, bbox_inches='tight')
print(f"✅ 同步到 copyparty: {copyparty_path}")
plt.close()