easy-tf-log 1.8

TensorFlow logging made easy

Easy TensorFlow Logging

Are you prototyping something and want to be able to magically graph some value without going through all the usual steps to set up TensorFlow logging properly?

easy-tf-log is a simple module to do just that.

from easy_tf_log import tflog

then you can do

for i in range(10):
    tflog('really_interesting_variable_name', i)

and you'll find a directory logs that you can point TensorBoard to

$ tensorboard --logdir logs

to get

Based on logging code from OpenAI's baselines.

Installation

pip install easy-tf-log

Note that TensorFlow must be installed separately.

Usage

By default, easy-tf-log saves event files to a directory logs. To change the directory, call easy_tf_log.set_dir(log_dir).

easy-tf-log also supports writing using an existing EventFileWriter created by e.g. an instance of tf.summary.FileWriter: call easy_tf_log.set_writer(file_writer.event_writer). (However, not that because EventsFileWriter uses a sub-thread to write events, this is not fork-safe. If you set this in one process and then try to use easy-tf-log a child process, it will hang.)

To log a value, use tflog(key, value). The step number for each key starts from zero and increments automatically. To set the step manually, specify the step argument.

See demo.py for a full demo.