python - 从 Keras 检查点加载

Question

我正在 Keras 中训练一个模型，我在其中使用以下代码保存了所有内容。

filepath = "project_model.hdh5"

checkpoint = ModelCheckpoint("project_model.hdf5", monitor='loss', verbose=1,
    save_best_only=False, mode='auto', period=1)

然后我使用以下代码来运行训练。

for _ in range(20):
    for j in range(len(mfcc_data_padded_transposed[j])):
        batch_input=[mfcc_data_padded_transposed[j]]
        batch_input = np.array(batch_input)
        batch_input = batch_input/np.max(batch_input)
        batch_output = [y_labels_mfcc[j]]
        batch_output = np.array(batch_output)
        input_lengths2 = input_lengths_mfcc[j]
        label_lengths2 = label_lengths_mfcc[j]
        input_lengths2 = np.array(input_lengths2)
        label_lengths2 = np.array(label_lengths2)
        inputs = {'the_input': batch_input,
         'the_labels': batch_output,
         'input_length': input_lengths2,
         'label_length': label_lengths2}
        outputs = {'ctc': np.zeros([1])} 
        model.fit(inputs, outputs, epochs=1, verbose =1, callbacks=[checkpoint])

我做了上面的检查点试验，因为我不确定我是否正确使用它。

现在，此训练以 0.001 的学习率完成。现在，在运行训练循环一段时间后，如果我决定将学习率更改为 .002，我是否必须运行与模型相关的所有代码(模型结构，然后是优化等)？并说我做到了，我如何从停止训练时的先前状态加载？另一个问题是，如果我重新启动 PC，并使用我之前在此处共享的检查点代码运行 jupyter 单元，这会替换之前保存的文件吗？加载保存的文件和权重并从那里恢复训练的理想方法是什么？我问的原因是因为当我遵循 Keras 文档时，它似乎只是从头开始。

Answer 1

Now after running the training loop for a while if I decide to change the learning rate to say .002, would I have to run all the codes that are related to the models (the model structure, then the optimization, etc)?

您可以在训练期间或加载模型后更新学习率。

请记住，学习率不属于模型架构，它属于优化器(在模型编译期间分配)。学习率是一个超参数，它调节梯度下降过程中权重更新的大小(在下面表示为 alpha):

因此，在初始训练后，您可以加载(保存的)模型，使用新的学习率更新优化器(并可能为编译器分配自定义对象)，然后继续训练。请记住，长时间训练模型后更改优化器本身可能会产生较差的准确度结果，因为您的模型现在必须重新校准以适应新优化器的权重计算。

how do I load from the previous state from when I stopped the training?

在 Keras 中，您可以选择保存/加载整个模型(包括架构、权重、优化器状态；或仅权重；或仅架构 (source)。

要保存/加载整个模型:

from keras.models import load_model

model.save('my_model.h5')
model = load_model('my_model.h5')

仅保存/加载模型权重:

model.save_weights('my_model_weights.h5')
model.load_weights('my_model_weights.h5')

您还可以在模型加载期间分配自定义对象:

model = load_model(filepath, custom_objects={'loss': custom_loss})

Other question is, if I restart the PC, and run the jupyter cell with checkpoint codes that I shared here earlier, would that replace the previously saved file?

取决于检查点中使用的文件路径:“如果文件路径是 weights.{epoch:02d}-{val_loss:.2f}.hdf5，则模型检查点将与 epoch 编号和验证损失一起保存在文件名”。因此，如果您对文件路径使用独特的格式，则可以避免覆盖以前保存的模型。 source

What is an ideal way to load the saved files and weights and resume training from there?

例子:

# Define model
model = keras.models.Sequential()

model.add(L.InputLayer([None],dtype='int32'))
model.add(L.Embedding(len(all_words),50))
model.add(keras.layers.Bidirectional(L.SimpleRNN(5,return_sequences=True)))

# Define softmax layer for every time step (hence TimeDistributed layer)
stepwise_dense = L.Dense(len(all_words),activation='softmax')
stepwise_dense = L.TimeDistributed(stepwise_dense)
model.add(stepwise_dense)

import keras.backend as K

# compile model with adam optimizer
model.compile('adam','categorical_crossentropy')

# print learning rate
print(f"Model learning rate is: {K.get_value(model.optimizer.lr):.3f}")

# train model
model.fit_generator(generate_batches(train_data), len(train_data)/BATCH_SIZE,
                    callbacks=[EvaluateAccuracy()], epochs=1)

# save model (weights, architecture, optimizer state)
model.save('my_model.h5')

# delete existing model
del model

结果

Model learning rate is: 0.001Epoch 1/11341/1343 [============================>.] - ETA: 0s - loss: 0.4288Measuring validation accuracy...Validation accuracy: 0.93138

from keras.models import load_model

# create new adam optimizer with le-04 learning rate (previous: 1e-03)
adam = keras.optimizers.Adam(lr=1e-4)

# load model
model = load_model('my_model.h5', compile=False)

# compile model and print new learning rate
model.compile(adam, 'categorical_crossentropy')
print(f"Model learning rate is: {K.get_value(model.optimizer.lr):.4f}")

# train model for 3 more epochs with new learning rate
print("Training model: ")
model.fit_generator(generate_batches(train_data),len(train_data)/BATCH_SIZE,
                    callbacks=[EvaluateAccuracy()], epochs=3,)

结果:

Model learning rate is: 0.0001Training model: Epoch 1/31342/1343 [============================>.] - ETA: 0s - loss: 0.0885Measuring validation accuracy...Validation accuracy: 0.935681344/1343 [==============================] - 41s - loss: 0.0885    Epoch 2/31342/1343 [============================>.] - ETA: 0s - loss: 0.0768Measuring validation accuracy...Validation accuracy: 0.939251344/1343 [==============================] - 39s - loss: 0.0768    Epoch 3/31343/1343 [============================>.] - ETA: 0s - loss: 0.0701Measuring validation accuracy...Validation accuracy: 0.94180

更多信息请访问 Keras FAQ针对具体情况。