flair.trainers.plugins.MetricRecord#

class flair.trainers.plugins.MetricRecord(name, value, global_step, typ, *, walltime=None)View on GitHub #

Bases: object

Represents a recorded metric value.

__init__(name, value, global_step, typ, *, walltime=None)View on GitHub #

Create a metric record.

Parameters:

name (Union[Iterable[str], str]) – Name of the metric.
typ (RecordType) – Type of metric.
value (Any) – Value of the metric (can be anything: scalar, tensor, image, etc.).
global_step (int) – The time_step of the log. This should be incremented the next time this metric is logged again. E.g. if you log every epoch, set the global_step to the current epoch.
walltime (Optional[float]) – Time of recording this metric.

Methods

`__init__`(name, value, global_step, typ, *[, ...])	Create a metric record.
`histogram`(name, value, global_step, *[, ...])
`is_type`(typ)
`scalar`(name, value, global_step, *[, walltime])
`scalar_list`(name, value, global_step, *[, ...])
`string`(name, value, global_step, *[, walltime])

Attributes

classmethod scalar(name, value, global_step, *, walltime=None)View on GitHub #

classmethod scalar_list(name, value, global_step, *, walltime=None)View on GitHub #

classmethod string(name, value, global_step, *, walltime=None)View on GitHub #

classmethod histogram(name, value, global_step, *, walltime=None)View on GitHub #