EarlyStopping¶

class paddle.callbacks. EarlyStopping ( monitor='loss', mode='auto', patience=0, verbose=1, min_delta=0, baseline=None, save_best_model=True ) [source]

Stop training when the given monitor stopped improving during evaluation by setting model.stop_training=True.

Parameters

• monitor (str) – Quantity to be monitored. Default: ‘loss’.

• mode (str|None) – Mode should be one of ‘auto’, ‘min’ or ‘max’. In ‘min’ mode, training will stop until monitored quantity stops decreasing. In ‘max’ mode, training will stop until monitored quantity stops increasing. In ‘auto’ mode, exact mode can be inferred by the name of monitor. If ‘acc’ in monitor, the mode will be considered as ‘max’, otherwise the mode will be set to ‘min’. Default: ‘auto’.

• patience (int) – Number of epochs with no improvement after which training will be stopped. Default: 0.

• verbose (int) – The verbosity mode, should be 0 or 1. When verbose=0, logs will not be printed. When verbose=1, logs will be printed. Default: 1.

• min_delta (int|float) – The minimum change of monitored quantity. If the change is less than min_delta, model could be considered as no improvement. Default: 0.

• baseline (int|float|None) – Baseline value for the monitored quantity. Training will stop if the model doesn’t show improvement over the baseline. Default: None.

• save_best_model (bool) – Whether to save best model. Default: True.

Examples

import paddle
from paddle import Model
from paddle.static import InputSpec
from paddle.vision.models import LeNet
from paddle.vision.datasets import MNIST
from paddle.metric import Accuracy
from paddle.nn import CrossEntropyLoss
import paddle.vision.transforms as T

device = paddle.set_device('cpu')
sample_num = 200
save_dir = './best_model_checkpoint'
transform = T.Compose(
    [T.Transpose(), T.Normalize([127.5], [127.5])])
train_dataset = MNIST(mode='train', transform=transform)
val_dataset = MNIST(mode='test', transform=transform)
net = LeNet()
optim = paddle.optimizer.Adam(
    learning_rate=0.001, parameters=net.parameters())

inputs = [InputSpec([None, 1, 28, 28], 'float32', 'x')]
labels = [InputSpec([None, 1], 'int64', 'label')]

model = Model(net, inputs=inputs, labels=labels)
model.prepare(
    optim,
    loss=CrossEntropyLoss(reduction="sum"),
    metrics=[Accuracy()])
callbacks = paddle.callbacks.EarlyStopping(
    'loss',
    mode='min',
    patience=1,
    verbose=1,
    min_delta=0,
    baseline=None,
    save_best_model=True)
model.fit(train_dataset,
          val_dataset,
          batch_size=64,
          log_freq=200,
          save_freq=10,
          save_dir=save_dir,
          epochs=20,
          callbacks=[callbacks])

on_train_begin ( logs=None )

on_train_begin¶

Called at the start of training.

Parameters

logs (dict) – The logs is a dict or None.

on_eval_end ( logs=None )

on_eval_end¶

Called at the end of evaluation.

Parameters

logs (dict) – The logs is a dict or None. The logs passed by paddle.Model is a dict contains ‘loss’, metrics and ‘batch_size’ of last batch of validation dataset.