TypeError: cannot pickle ‘_thread.lock‘ objec M2 Mac笔记本出现的问题

我在训练一个深度学习模型的时候出现了这个错误，但是在colab上可以运行。

具体代码如下：

def run_model():
    ### Open a strategy scope.
    with strategy.scope():
        # Everything that creates variables should be under the strategy scope.
        # In general this is only model construction & `compile()`.
        model = get_compiled_model()
    ### Set train steps and validation steps
    train_steps =  len(train_generator.labels)/ batch_size
    val_steps = len(validation_generator.labels) / batch_size
    
    #### set the path to save models having lowest validation loss during training
    save_model_dir = './models/'
    if not os.path.exists(save_model_dir):
        os.mkdir(save_model_dir)
    filepath= "models/breast-"+args.structure+"-fold" + str(i+1) + "-" + database + "-" + args.model_name + "-" + str(image_size) + "-" + str(batch_size) + "-"+str(args.lr)+ ".h5"   
    
    
    checkpoint = ModelCheckpoint(filepath, monitor='val_loss', verbose=1,save_best_only=True, mode='min')
    history = model.fit(
            train_generator,
            epochs=num_epoches,
            steps_per_epoch=train_steps,
            validation_data=validation_generator,
            validation_steps=val_steps,
            use_multiprocessing=False,
            workers=10,
            callbacks=[checkpoint])
    ### Save training loss
    train_auc = history.history['auc']
    val_auc = history.history['val_auc']
    train_loss = history.history['loss']
    val_loss = history.history['val_loss']
    d_loss = pd.DataFrame({'train_auc':train_auc, 'val_auc':val_auc, 'train_loss':train_loss, 'val_loss':val_loss})
    save_loss_dir = './loss'
    if not os.path.exists(save_loss_dir):
        os.mkdir(save_loss_dir)
    d_loss.to_csv("loss/breast-"+args.structure+"-fold" + str(i+1) + "-" + database + "-" +  args.model_name + "-" + str(image_size) + "-" + str(batch_size) + "-"+str(args.lr)+ ".csv", index=False)

    
# In[17]:
for i in range(5):
    #df_train=pd.read_csv("dataframe/breast_train_fold"+str(i+1)+".csv")
    #df_val=pd.read_csv("dataframe/breast_val_fold"+str(i+1)+".csv")
    df_train=pd.read_csv("fold"+str(i)+"_train.csv")
    df_val=pd.read_csv("fold"+str(i)+"_test.csv")
    train_data_generator = ImageDataGenerator(
                                    rescale=1./255,
                                    preprocessing_function=preprocess_input,
                                    rotation_range=10,
                                    width_shift_range=0.1,
                                    height_shift_range=0.1,
                                    shear_range=0.1,
                                    zoom_range=0.1,
                                    horizontal_flip=True,
                                    fill_mode='nearest'
                                   )
    data_generator = ImageDataGenerator(rescale=1./255, preprocessing_function=preprocess_input)
    train_generator = train_data_generator.flow_from_dataframe(
        dataframe=df_train,
        x_col = 'filename',
        y_col = 'label',
        target_size=(image_size, image_size),
        batch_size=batch_size,
        shuffle=True,
        seed=726,
        class_mode='categorical',
       	)
    validation_generator = data_generator.flow_from_dataframe(
        dataframe=df_val,
        x_col = 'filename',
        y_col = 'label',
        target_size=(image_size, image_size),
        batch_size=batch_size,
        shuffle=True,
        seed=726,
        class_mode='categorical',
        )
    num_classes =len(train_generator.class_indices)
    run_model()

当时我以为是我tensflow版本不对，但我发现colab的版本也就2.12

所以我谷歌了一下发现有人提到use_multiprocessing更改为False，但是我以为是将train_data_generator.flow_from_datafram里添加，没注意到

history = model.fit(
            train_generator,
            epochs=num_epoches,
            steps_per_epoch=train_steps,
            validation_data=validation_generator,
            validation_steps=val_steps,
            use_multiprocessing=False,
            workers=10,
            callbacks=[checkpoint])
            

model.fit本身就有use_multiprocessing=False，false是我修正过的，所以目前代码可以在笔记本上运行了。

顺带说一下，m2芯片还是比较强的，但长时间工作那个风扇转速感觉CPU要烧了，训练个小项目还是比较方便。