我如何像在Keras中一样详细地在tensorflow中打印ETA

时间:2018-08-10 09:39:16

标签: python tensorflow keras epoch

我需要估算每个时期在Tensorflow中训练我的模型所需的估计时间。在keras中,当我用冗长的方式预测类时,我得到的是这样的:

Epoch 1/70 17800/34960 [==============>...............] - ETA: 22s - 
loss: 7.3802 - acc: 0.0631
34960/34960 [==============================] - 44s 1ms/step - loss: 7.1693 - 
acc: 0.0641 
Epoch 2/70 34960/34960 [==============================] - 42s 
1ms/step - loss: 6.7803 - acc: 0.0646
Epoch 3/70 34960/34960 
[==============================] - 43s 1ms/step - loss: 6.6338 - acc: 0.0678 
Epoch 4/70 5200/34960 [===>..........................] - ETA: 36s - loss: 
6.3998 - acc: 0.0737
34960/34960 [==============================] - 43s 1ms/step - loss: 6.4653 - 
acc: 0.0753 
Epoch 5/70 34960/34960 [==============================] - 43s 

如何仅在Tensorflow中获得相同类型的信息。当我打印(tf.logging.set_verbosity(tf.logging.INFO)时,它显示为“无”

0 个答案:

没有答案