Question

我从未使用过检查点，所以我遵循了我发现的教程，它似乎有效，但在第一个纪元之后，它给了我一个奇怪的错误。我实际上在使用 google colab。这是我的代码：

model.compile(loss='mae',
                optimizer='adam'
              )

filepath="weights.best.hdf5"
checkpoint = ModelCheckpoint(filepath, monitor='val_loss', verbose=1, save_best_only=True, mode='max')
callbacks_list = [checkpoint]

model.fit([train_xw[192:], train_xl],train_yl,
            validation_data=([test_xw[192:] ,test_xl], test_yl),
            batch_size=batch_size,
            callbacks=[callbacks_list],
            verbose = verbose,
            epochs = epochs)

错误如下

Epoch 1/150
816/816 [==============================] - 8s 8ms/step - loss: 125.2619 - mae: 125.2619 - val_loss: 136.2924 - val_mae: 136.2924

Epoch 00001: val_loss improved from -inf to 136.29237, saving model to weights.best.hdf5
---------------------------------------------------------------------------
KeyError                                  Traceback (most recent call last)
<ipython-input-116-835d797d0cff> in <module>()
     13             callbacks=[callbacks_list],
     14             verbose = verbose,
---> 15             epochs = epochs)

9 frames
/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/engine/functional.py in get_network_config(network, serialize_layer_fn)
   1366     if node_key not in network._network_nodes:
   1367       continue
-> 1368     new_node_index = node_conversion_map[node_key]
   1369     model_inputs.append(
   1370         tf_utils.ListWrapper([layer.name, new_node_index, tensor_index]))

KeyError: 'input_1_ib-0'

尝试在神经网络中保存检查点时出错？

0 个答案: