Attribute types reference

An attribute is the location of a piece of metadata in a Neptune run object.

note

In Neptune 2.x, attributes are called fields.

The metadata type and logging method together determine the resulting attribute type. The type determines the operations available for the attribute.

Metadata type	Example	Logging method	Resulting attribute type
Single value	Parameters, final scores, text, time	log_configs()	Float, Integer, Boolean, String, or Datetime
Series	Metrics, loss, accuracy	log_metrics()	FloatSeries
Tags	Text tags to annotate runs or assign them to groups	add_tags()	StringSet

Related

• Namespaces and attributes

Simple types

Simple attribute types can represent model configurations, scores, or other single values. The following types are available:

• Boolean
• Datetime
• Float
• Integer
• String

To assign single values to a Neptune run, use the log_configs() method. To learn more, see Log parameters and model configuration.

You can display simple attribute:

• as columns in the experiments table
• in Single value widgets

Series

A series attribute collects a sequence of values into a single attribute. You create a series with the log_metrics() function. Values are added to a series attribute iteratively: each call adds a new value to the sequence.

The following Series types are supported:

Attribute type	How to display
FloatSeries	Experiments table, Charts tab, or Chart widget

To learn more, see Log metrics.

---

Boolean

Attribute type representing a Boolean value.

Example

with Run(...) as run:
    run.log_configs(
        {"params/use_preprocessing": True}
    )

---

Datetime

Attribute type representing a datetime value.

Example

Logging custom training end time

from datetime import datetime

with Run(...) as run:
    run.log_configs(
        {"train/end": datetime.now()}
    )

Logging other time-related metadata

from datetime import datetime

with Run(...) as run:
    run.log_configs(
        {"dataset/creation_date": datetime.fromisoformat("1998-11-01")}
    )

---

Float

Attribute type representing a floating point value.

Example

run = Run()
run.log_configs(
    {"params/learning_rate": 0.001}
)

---

FloatSeries

Attribute containing a series of float values, for example:

• Training metrics
• Change of performance of the model in production

You can index the series by step or by time.

Example

run = Run()
for step in epoch:
    # your training loop
    run.log_metrics(
        data={"loss": 0.14},
        step=step,
    )

To learn more, see Log metrics.

---

Integer

Attribute type representing an integer.

Example

with Run(...) as run:
    run.log_configs(
        {"params/max_epochs": 100}
    )

---

String

Attribute type representing a string.

Example

with Run(...) as run:
    run.log_configs(
        {"params/optimizer": "Adam"}
    )

note

Due to a technical limitation, only 1000 characters are indexed in String attributes. This means that when searching the experiments table, only the first 1000 characters are considered.

---

StringSet

An attribute containing an unorganized set of strings.

The supported StringSet attributes are sys/tags and sys/group_tags. You can't create custom StringSet attribute paths.

Examples

Adding tags creates a StringSet at the path sys/tags:

with Run(...) as run:
    run.add_tags(["spaces are ok", "v1.2.3"])

Adding group tags creates a StringSet at the path sys/group_tags:

with Run(...) as run:
    run.add_tags(tags=["albino", "parent"], group_tags=True)

Related

• Adding tags
• Groups