我正在使用1 - dice_coefficient作为损失指标，在使用tensorflow 2.0进行训练时，接下来的冗长内容使我感到荒谬的损失，目前我在每个conv2d层都引入了活动调节器（l2正则化）。
损失计算正确，为什么冗长不正确？
Epoch 1/10
loss: 0.6226288 dice: 0.37737122
1/32057 [..............................] - ETA: 164:21:12 - loss: 58262.4844 - dice_coef: 0.3774
loss: 0.66217625 dice: 0.33782378
2/32057 [..............................] - ETA: 92:42:14 - loss: 47621.1680 - dice_coef: 0.3576
loss: 0.7298874 dice: 0.27011257
3/32057 [..............................] - ETA: 68:58:37 - loss: 41726.3203 - dice_coef: 0.3284
loss: 0.91065186 dice: 0.089348145
4/32057 [..............................] - ETA: 57:40:56 - loss: 36527.0034 - dice_coef: 0.2687
loss: 0.6360147 dice: 0.36398527