本文整理汇总了Python中tensorflow.keras.callbacks方法的典型用法代码示例。如果您正苦于以下问题：Python keras.callbacks方法的具体用法？Python keras.callbacks怎么用？Python keras.callbacks使用的例子？那么, 这里精选的方法代码示例或许可以为您提供帮助。您也可以进一步了解该方法所在类tensorflow.keras的用法示例。
在下文中一共展示了keras.callbacks方法的1个代码示例，这些例子默认根据受欢迎程度排序。您可以为喜欢或者感觉有用的代码点赞，您的评价将有助于系统推荐出更棒的Python代码示例。
示例1: init_callback_objects
# 需要导入模块: from tensorflow import keras [as 别名]
# 或者: from tensorflow.keras import callbacks [as 别名]
def init_callback_objects(callbacks, logger):
"""
Initialize a list of tf.keras/custom callback descriptors.
Args:
callbacks: List of callback descriptions. Each list entry should be
either a dictionary of the format
{"class_name": <name_of_CB_class>,
"kwargs": {<dict_with_kwargs>}}
... or an already initialized callback object
(which will be skipped).
logger: A mpunet logger object logging to screen and/or file
Returns:
A list of initialized callbacks
"""
from mpunet import callbacks as tcb
cb_objs = []
cb_dict = {}
for i, callback in enumerate(callbacks):
if not isinstance(callback, dict):
# CB already initialized
cb = callback
kwargs = {"params": "?"}
cls_name = callback.__class__.__name__
start_from = 0
else:
kwargs = callback["kwargs"]
cls_name = callback["class_name"]
start_from = callback.get("start_from")
if callback.get("pass_logger"):
kwargs["logger"] = logger
try:
cb = getattr(tfcb, cls_name, None) or getattr(tcb, cls_name)
except AttributeError as e:
raise ValueError("No callback named %s" % cls_name) from e
cb = cb(**kwargs)
if start_from:
logger("OBS: '%s' activates at epoch %i" % (cls_name, start_from))
cb = DelayedCallback(callback=cb,
start_from=start_from,
logger=logger)
cb_objs.append(cb)
cb_dict[cls_name] = cb
logger("[%i] Using callback: %s(%s)" % (i+1, cb.__class__.__name__,
", ".join(["%s=%s" % (a, kwargs[a]) for a in kwargs])))
return cb_objs, cb_dict