老师，下午好，请教一个问题？
本节中，我照着视频敲的代码。

model = keras.models.Sequential([
    keras.layers.Dense(30, activation='relu', 
                       input_shape=x_train.shape[1:]), # x_train.shape[1:] 即 x_train.shape[1:8]
    keras.layers.Dense(1) 
])
model.summary() 
model.compile(loss="mean_squared_error", optimizer="sgd")
callbacks = [keras.callbacks.EarlyStopping(patience=5, min_delta=1e-3)]

下面是执行fit()的结果：

Epoch 1/100
363/363 [==============================] - 1s 2ms/step - loss: 3.5061 - val_loss: 0.4793
Epoch 2/100
363/363 [==============================] - 0s 877us/step - loss: 0.4930 - val_loss: 0.4187
Epoch 3/100
363/363 [==============================] - 0s 890us/step - loss: 0.3834 - val_loss: 0.4354
Epoch 4/100
363/363 [==============================] - 0s 863us/step - loss: 0.4189 - val_loss: 0.3943
Epoch 5/100
363/363 [==============================] - 0s 905us/step - loss: 0.3898 - val_loss: 0.3839
Epoch 6/100
363/363 [==============================] - 0s 872us/step - loss: 0.3588 - val_loss: 0.3770
Epoch 7/100
363/363 [==============================] - 0s 891us/step - loss: 0.3778 - val_loss: 0.4224
Epoch 8/100
363/363 [==============================] - 0s 872us/step - loss: 0.3715 - val_loss: 0.3769
Epoch 9/100
363/363 [==============================] - 0s 887us/step - loss: 0.3555 - val_loss: 0.3731
Epoch 10/100
363/363 [==============================] - 0s 921us/step - loss: 0.3442 - val_loss: 0.3687
Epoch 11/100
363/363 [==============================] - 0s 875us/step - loss: 0.3628 - val_loss: 0.3761
Epoch 12/100
363/363 [==============================] - 0s 883us/step - loss: 0.3781 - val_loss: 0.3660
Epoch 13/100
363/363 [==============================] - 0s 890us/step - loss: 0.3531 - val_loss: 0.3649
Epoch 14/100
363/363 [==============================] - 0s 945us/step - loss: 0.3629 - val_loss: 0.3672
Epoch 15/100
363/363 [==============================] - 0s 914us/step - loss: 0.3557 - val_loss: 0.3653
Epoch 16/100
363/363 [==============================] - 0s 1ms/step - loss: 0.3558 - val_loss: 0.3618
Epoch 17/100
363/363 [==============================] - 0s 905us/step - loss: 0.3498 - val_loss: 0.3611
Epoch 18/100
363/363 [==============================] - 0s 912us/step - loss: 0.3538 - val_loss: 0.3593
Epoch 19/100
363/363 [==============================] - 0s 886us/step - loss: 0.3612 - val_loss: 0.3971
Epoch 20/100
363/363 [==============================] - 0s 909us/step - loss: 0.3915 - val_loss: 0.3673
Epoch 21/100
363/363 [==============================] - 0s 901us/step - loss: 0.3549 - val_loss: 0.3646
Epoch 22/100
363/363 [==============================] - 0s 864us/step - loss: 0.3437 - val_loss: 0.3587
Epoch 23/100
363/363 [==============================] - 0s 900us/step - loss: 0.3506 - val_loss: 0.3593

我反复看了，在epoch 16/100 , 17/100 ， val_loss 的差为 0.3618 - 0.3611 = 0.0007 < 1e-3 , 还有一次是
epoch 22/100, 23/100 ， val_loss的差为 0.3587 - 0.3593 = -0.0006 < 1e-3,
我仔细检查了很多遍，
请问老师，
问题1： 具体是哪5次呢？（负数应该不算吧，因为我数过，负数算的话就不止5次了。）
问题2： 上面的计算结果 -0.0006 要取绝对值吗？

谢谢老师指导！