在一个数据集上训练模型,训练很多个epoch后,学习率lr已经变得很小,需要改变lr
想在新数据集上加载此模型,使用model.load_weights后,模型的学习率仍为原来的小学习率
加入代码
from keras import backend as K
# To get learning rate
print(K.get_value(model.optimizer.lr))
# To set learning rate
K.set_value(model.optimizer.lr, 0.001)
keras.__version__ # 2.0.2
我的原始代码:
weightsPath = '/content/drive/MyDrive/Violence/Datasets/save_path/RealLife/bestValPath/'
model.load_weights(f'{weightsPath}').expect_partial()
optimizer = Adam(lr=5e-04,amsgrad=True)
model.compile(optimizer=optimizer, loss='binary_crossentropy', metrics=['acc'])
使用此代码学习率无法改变
改后代码:
weightsPath = '/content/drive/MyDrive/Violence/Datasets/save_path/RealLife/bestValPath/'
model.load_weights(f'{weightsPath}').expect_partial()
optimizer = Adam(lr=5e-04,amsgrad=True)
model.compile(optimizer=optimizer, loss='binary_crossentropy', metrics=['acc'])
K.set_value(model.optimizer.lr,5e-04)
注意不能删除这两行代码
optimizer = Adam(lr=5e-04,amsgrad=True)
model.compile(optimizer=optimizer, loss='binary_crossentropy', metrics=['acc'])
文章出处登录后可见!
已经登录?立即刷新