jupyter 中tf 搭建神经网络 如何确定自己的GPU参与了计算？

Question

采用mnist的数据集练习 神经网络分类问题，搭建完模型后的fit过程如下：

%%time
model.fit(x_train_reshaped,y_train,epochs = 10,batch_size = 64,
          validation_data = (x_valid_reshaped,y_valid))
# 这里面一定要用reshaped的数据才行

运算结果如下：

Epoch 1/10
938/938 [==============================] - 4s 4ms/step - loss: 3.0898 - sparse_categorical_accuracy: 0.6319 - val_loss: 0.7920 - val_sparse_categorical_accuracy: 0.7922
Epoch 2/10
938/938 [==============================] - 3s 4ms/step - loss: 0.6585 - sparse_categorical_accuracy: 0.8172 - val_loss: 0.5834 - val_sparse_categorical_accuracy: 0.8415
Epoch 3/10
938/938 [==============================] - 3s 4ms/step - loss: 0.5044 - sparse_categorical_accuracy: 0.8641 - val_loss: 0.4774 - val_sparse_categorical_accuracy: 0.8803
Epoch 4/10
938/938 [==============================] - 3s 4ms/step - loss: 0.4253 - sparse_categorical_accuracy: 0.8852 - val_loss: 0.3929 - val_sparse_categorical_accuracy: 0.8931
Epoch 5/10
938/938 [==============================] - 4s 4ms/step - loss: 0.3789 - sparse_categorical_accuracy: 0.8986 - val_loss: 0.3681 - val_sparse_categorical_accuracy: 0.9028
Epoch 6/10
938/938 [==============================] - 4s 4ms/step - loss: 0.3455 - sparse_categorical_accuracy: 0.9064 - val_loss: 0.3584 - val_sparse_categorical_accuracy: 0.9103
Epoch 7/10
938/938 [==============================] - 3s 4ms/step - loss: 0.3146 - sparse_categorical_accuracy: 0.9131 - val_loss: 0.3123 - val_sparse_categorical_accuracy: 0.9160
Epoch 8/10
938/938 [==============================] - 3s 4ms/step - loss: 0.2925 - sparse_categorical_accuracy: 0.9198 - val_loss: 0.3075 - val_sparse_categorical_accuracy: 0.9169
Epoch 9/10
938/938 [==============================] - 4s 4ms/step - loss: 0.2782 - sparse_categorical_accuracy: 0.9214 - val_loss: 0.3440 - val_sparse_categorical_accuracy: 0.9157
Epoch 10/10
938/938 [==============================] - 4s 4ms/step - loss: 0.2652 - sparse_categorical_accuracy: 0.9247 - val_loss: 0.3154 - val_sparse_categorical_accuracy: 0.9147
Wall time: 35.9 s

我想问在这个过程中GPU 有没有参与计算？ 如果没有，怎样设置才可以GPU 参与计算，在这之前我也将CUDA 显卡驱动什么的都装好了。
数据量测试集是 60000784 验证集是10000784.

Answer 1

有个简单的办法，提前打开任务管理器翻到GPU那一栏，在训练或者预测数据的时候观察各项指标特别是复制有没有起伏