jactorch.train.tb

Classes

TBGroupMeters	A group of meters that can be updated and logged to tensorboard.
TBLogger	A simple wrapper for tensorboard.

Class TBGroupMeters

class TBGroupMeters

Bases: GroupMeters

A group of meters that can be updated and logged to tensorboard.

__init__(tb_logger)

__new__(**kwargs)

dump(filename, values='avg')

flush()

format(caption, values, kv_format, glue)

format_simple(caption=None, values='avg', compressed=True)

items()

reset()

update(updates=None, value=None, n=1, prefix=None, **kwargs)

Update the meters.

Example

>>> meters.update(key, value)
>>> meters.update({key1: value1, key2: value2})
>>> meters.update(key1=value1, key2=value2)

property avg

property count

property std

property sum

property val

Class TBLogger

class TBLogger

Bases: object

A simple wrapper for tensorboard.

Source: https://raw.githubusercontent.com/SherlockLiao/pytorch-beginner/

__init__(log_dir)

Creates a summary writer logging to log_dir.

Parameters:

log_dir (str) – the directory to save the logs.

__new__(**kwargs)

flush()

histo_summary(tag, values, step, bins=1000)

image_summary(tag, images, step)

scalar_summary(tag, value, step)