1. tensorflow log Introduction
TensorFlow use five different levels of log messages. Ascending order, they are DEBUG
, INFO
, WARN
, ERROR
and FATAL
. When you record at any of these levels configuration log, TensorFlow output all log messages corresponding to the level greater than that level and all log. For example, if the ERROR logging level is set, and you will receive a log output and a FATAL ERROR message contains, if a set DEBUG level, will obtain all log messages from the five levels. By default, Tensorflow in the WARN logging level configuration, but when tracking model training, you need to adjust the level INFO, which will provide feedback for other operations in progress.
2. tf.logging.set_verbosity()
Setting a threshold for which log messages will be recorded. Python log corresponding to level
the parameter
# 设置日志级别,此时显示大于等于INFO级别的日志
tf.logging.set_verbosity(tf.logging.INFO)
3. Logging
tf.logging.info("hellow,word") # 记录一条info级别的日志
tf.logging.debug("hellow,word")
tf.logging.warn("helllow, word")
tf.logging.error("hellow, word")
tf.logging.fatal("hellow, word")
Example 4
Under 4.1 default default, Tensorflow be configured WARN logging level
import tensorflow as tf
tf.logging.info("hellow word, info level")
tf.logging.debug("hellow word, debug level")
tf.logging.warn("hellow word, warn level")
tf.logging.error("hellow word, error level")
tf.logging.fatal("hellow word, fatal level")
Export
W0208 12:12:14.905251 90696 test.py:5] hellow word, warn level
E0208 12:12:14.905251 90696 test.py:6] hellow word, error level
E0208 12:12:14.906248 90696 test.py:7] CRITICAL - hellow word, fatal level
[Finished in 49.2s]
4.2 Setting level logging
Info output level greater than or equal to log
import tensorflow as tf
tf.logging.set_verbosity(tf.logging.INFO)
tf.logging.info("hellow word, info level")
tf.logging.debug("hellow word, debug level")
tf.logging.warn("hellow word, warn level")
tf.logging.error("hellow word, error level")
tf.logging.fatal("hellow word, fatal level")
Export
I0208 12:15:57.261396 69356 test.py:3] hellow word, info level
W0208 12:15:57.261396 69356 test.py:5] hellow word, warn level
E0208 12:15:57.261396 69356 test.py:6] hellow word, error level
E0208 12:15:57.261396 69356 test.py:7] CRITICAL - hellow word, fatal level
5. Other matters
5.1 tf.logging.info(msg, *args, **args)
Parameter args
is with msg placeholder used, such as
tf.logging.info("now %d epochs on %s ", 100, "test.py")
Export
I0208 12:36:35.534926 86200 test.py:3] now 100 epochs on test.py